package nomad import ( "errors" "fmt" "reflect" "strings" "testing" "time" memdb "github.com/hashicorp/go-memdb" msgpackrpc "github.com/hashicorp/net-rpc-msgpackrpc" "github.com/hashicorp/raft" "github.com/kr/pretty" "github.com/stretchr/testify/assert" "github.com/stretchr/testify/require" "github.com/hashicorp/nomad/acl" "github.com/hashicorp/nomad/helper" "github.com/hashicorp/nomad/helper/uuid" "github.com/hashicorp/nomad/nomad/mock" "github.com/hashicorp/nomad/nomad/structs" "github.com/hashicorp/nomad/testutil" ) func TestJobEndpoint_Register(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } serviceName := out.TaskGroups[0].Tasks[0].Services[0].Name expectedServiceName := "web-frontend" if serviceName != expectedServiceName { t.Fatalf("Expected Service Name: %s, Actual: %s", expectedServiceName, serviceName) } // Lookup the evaluation eval, err := state.EvalByID(ws, resp.EvalID) if err != nil { t.Fatalf("err: %v", err) } if eval == nil { t.Fatalf("expected eval") } if eval.CreateIndex != resp.EvalCreateIndex { t.Fatalf("index mis-match") } if eval.Priority != job.Priority { t.Fatalf("bad: %#v", eval) } if eval.Type != job.Type { t.Fatalf("bad: %#v", eval) } if eval.TriggeredBy != structs.EvalTriggerJobRegister { t.Fatalf("bad: %#v", eval) } if eval.JobID != job.ID { t.Fatalf("bad: %#v", eval) } if eval.JobModifyIndex != resp.JobModifyIndex { t.Fatalf("bad: %#v", eval) } if eval.Status != structs.EvalStatusPending { t.Fatalf("bad: %#v", eval) } if eval.CreateTime == 0 { t.Fatalf("eval CreateTime is unset: %#v", eval) } if eval.ModifyTime == 0 { t.Fatalf("eval ModifyTime is unset: %#v", eval) } } func TestJobEndpoint_Register_PreserveCounts(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.TaskGroups[0].Name = "group1" job.TaskGroups[0].Count = 10 job.Canonicalize() // Register the job require.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, }, &structs.JobRegisterResponse{})) // Check the job in the FSM state state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.NoError(err) require.NotNil(out) require.Equal(10, out.TaskGroups[0].Count) // New version: // new "group2" with 2 instances // "group1" goes from 10 -> 0 in the spec job = job.Copy() job.TaskGroups[0].Count = 0 // 10 -> 0 in the job spec job.TaskGroups = append(job.TaskGroups, job.TaskGroups[0].Copy()) job.TaskGroups[1].Name = "group2" job.TaskGroups[1].Count = 2 // Perform the update require.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", &structs.JobRegisterRequest{ Job: job, PreserveCounts: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, }, &structs.JobRegisterResponse{})) // Check the job in the FSM state out, err = state.JobByID(nil, job.Namespace, job.ID) require.NoError(err) require.NotNil(out) require.Equal(10, out.TaskGroups[0].Count) // should not change require.Equal(2, out.TaskGroups[1].Count) // should be as in job spec } func TestJobEndpoint_Register_EvalPriority(t *testing.T) { t.Parallel() requireAssert := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.TaskGroups[0].Name = "group1" job.Canonicalize() // Register the job. requireAssert.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, EvalPriority: 99, }, &structs.JobRegisterResponse{})) // Grab the eval from the state, and check its priority is as expected. state := s1.fsm.State() out, err := state.EvalsByJob(nil, job.Namespace, job.ID) requireAssert.NoError(err) requireAssert.Len(out, 1) requireAssert.Equal(99, out[0].Priority) } func TestJobEndpoint_Register_Connect(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.TaskGroups[0].Tasks[0].Services = nil job.TaskGroups[0].Networks = structs.Networks{{ Mode: "bridge", }} job.TaskGroups[0].Services = []*structs.Service{{ Name: "backend", PortLabel: "8080", Connect: &structs.ConsulConnect{ SidecarService: &structs.ConsulSidecarService{}, }}, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse require.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) require.NotZero(resp.Index) // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) require.NoError(err) require.NotNil(out) require.Equal(resp.JobModifyIndex, out.CreateIndex) // Check that the sidecar task was injected require.Len(out.TaskGroups[0].Tasks, 2) sidecarTask := out.TaskGroups[0].Tasks[1] require.Equal("connect-proxy-backend", sidecarTask.Name) require.Equal("connect-proxy:backend", string(sidecarTask.Kind)) require.Equal("connect-proxy-backend", out.TaskGroups[0].Networks[0].DynamicPorts[0].Label) // Check that round tripping the job doesn't change the sidecarTask out.Meta["test"] = "abc" req.Job = out require.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) require.NotZero(resp.Index) // Check for the new node in the FSM state = s1.fsm.State() ws = memdb.NewWatchSet() out, err = state.JobByID(ws, job.Namespace, job.ID) require.NoError(err) require.NotNil(out) require.Equal(resp.JobModifyIndex, out.CreateIndex) require.Len(out.TaskGroups[0].Tasks, 2) require.Exactly(sidecarTask, out.TaskGroups[0].Tasks[1]) } func TestJobEndpoint_Register_ConnectIngressGateway_minimum(t *testing.T) { t.Parallel() r := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // job contains the minimalist possible gateway service definition job := mock.ConnectIngressGatewayJob("host", false) // Create the register request req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse r.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) r.NotZero(resp.Index) // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) r.NoError(err) r.NotNil(out) r.Equal(resp.JobModifyIndex, out.CreateIndex) // Check that the gateway task got injected r.Len(out.TaskGroups[0].Tasks, 1) task := out.TaskGroups[0].Tasks[0] r.Equal("connect-ingress-my-ingress-service", task.Name) r.Equal("connect-ingress:my-ingress-service", string(task.Kind)) r.Equal("docker", task.Driver) r.NotNil(task.Config) // Check the CE fields got set service := out.TaskGroups[0].Services[0] r.Equal(&structs.ConsulIngressConfigEntry{ TLS: nil, Listeners: []*structs.ConsulIngressListener{{ Port: 2000, Protocol: "tcp", Services: []*structs.ConsulIngressService{{ Name: "service1", }}, }}, }, service.Connect.Gateway.Ingress) // Check that round-tripping does not inject a duplicate task out.Meta["test"] = "abc" req.Job = out r.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) r.NotZero(resp.Index) // Check for the new node in the fsm state = s1.fsm.State() ws = memdb.NewWatchSet() out, err = state.JobByID(ws, job.Namespace, job.ID) r.NoError(err) r.NotNil(out) r.Equal(resp.JobModifyIndex, out.CreateIndex) // Check we did not re-add the task that was added the first time r.Len(out.TaskGroups[0].Tasks, 1) } func TestJobEndpoint_Register_ConnectIngressGateway_full(t *testing.T) { t.Parallel() r := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // reconfigure job to fill in all the possible fields job := mock.ConnectIngressGatewayJob("bridge", false) job.TaskGroups[0].Services[0].Connect = &structs.ConsulConnect{ Gateway: &structs.ConsulGateway{ Proxy: &structs.ConsulGatewayProxy{ ConnectTimeout: helper.TimeToPtr(1 * time.Second), EnvoyGatewayBindTaggedAddresses: true, EnvoyGatewayBindAddresses: map[string]*structs.ConsulGatewayBindAddress{ "service1": { Address: "10.0.0.1", Port: 2001, }, "service2": { Address: "10.0.0.2", Port: 2002, }, }, EnvoyGatewayNoDefaultBind: true, Config: map[string]interface{}{ "foo": 1, "bar": "baz", }, }, Ingress: &structs.ConsulIngressConfigEntry{ TLS: &structs.ConsulGatewayTLSConfig{ Enabled: true, }, Listeners: []*structs.ConsulIngressListener{{ Port: 3000, Protocol: "tcp", Services: []*structs.ConsulIngressService{{ Name: "db", }}, }, { Port: 3001, Protocol: "http", Services: []*structs.ConsulIngressService{{ Name: "website", Hosts: []string{"10.0.1.0", "10.0.1.0:3001"}, }}, }}, }, }, } // Create the register request req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse r.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) r.NotZero(resp.Index) // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) r.NoError(err) r.NotNil(out) r.Equal(resp.JobModifyIndex, out.CreateIndex) // Check that the gateway task got injected r.Len(out.TaskGroups[0].Tasks, 1) task := out.TaskGroups[0].Tasks[0] r.Equal("connect-ingress-my-ingress-service", task.Name) r.Equal("connect-ingress:my-ingress-service", string(task.Kind)) r.Equal("docker", task.Driver) r.NotNil(task.Config) // Check that the ingress service is all set service := out.TaskGroups[0].Services[0] r.Equal("my-ingress-service", service.Name) r.Equal(&structs.ConsulIngressConfigEntry{ TLS: &structs.ConsulGatewayTLSConfig{ Enabled: true, }, Listeners: []*structs.ConsulIngressListener{{ Port: 3000, Protocol: "tcp", Services: []*structs.ConsulIngressService{{ Name: "db", }}, }, { Port: 3001, Protocol: "http", Services: []*structs.ConsulIngressService{{ Name: "website", Hosts: []string{"10.0.1.0", "10.0.1.0:3001"}, }}, }}, }, service.Connect.Gateway.Ingress) // Check that round-tripping does not inject a duplicate task out.Meta["test"] = "abc" req.Job = out r.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) r.NotZero(resp.Index) // Check for the new node in the fsm state = s1.fsm.State() ws = memdb.NewWatchSet() out, err = state.JobByID(ws, job.Namespace, job.ID) r.NoError(err) r.NotNil(out) r.Equal(resp.JobModifyIndex, out.CreateIndex) // Check we did not re-add the task that was added the first time r.Len(out.TaskGroups[0].Tasks, 1) } func TestJobEndpoint_Register_ConnectExposeCheck(t *testing.T) { t.Parallel() r := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Setup the job we are going to register job := mock.Job() job.TaskGroups[0].Tasks[0].Services = nil job.TaskGroups[0].Networks = structs.Networks{{ Mode: "bridge", DynamicPorts: []structs.Port{{ Label: "hcPort", To: -1, }, { Label: "v2Port", To: -1, }}, }} job.TaskGroups[0].Services = []*structs.Service{{ Name: "backend", PortLabel: "8080", Checks: []*structs.ServiceCheck{{ Name: "check1", Type: "http", Protocol: "http", Path: "/health", Expose: true, PortLabel: "hcPort", Interval: 1 * time.Second, Timeout: 1 * time.Second, }, { Name: "check2", Type: "script", Command: "/bin/true", TaskName: "web", Interval: 1 * time.Second, Timeout: 1 * time.Second, }, { Name: "check3", Type: "grpc", Protocol: "grpc", Path: "/v2/health", Expose: true, PortLabel: "v2Port", Interval: 1 * time.Second, Timeout: 1 * time.Second, }}, Connect: &structs.ConsulConnect{ SidecarService: &structs.ConsulSidecarService{}}, }} // Create the register request req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse r.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) r.NotZero(resp.Index) // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) r.NoError(err) r.NotNil(out) r.Equal(resp.JobModifyIndex, out.CreateIndex) // Check that the new expose paths got created r.Len(out.TaskGroups[0].Services[0].Connect.SidecarService.Proxy.Expose.Paths, 2) httpPath := out.TaskGroups[0].Services[0].Connect.SidecarService.Proxy.Expose.Paths[0] r.Equal(structs.ConsulExposePath{ Path: "/health", Protocol: "http", LocalPathPort: 8080, ListenerPort: "hcPort", }, httpPath) grpcPath := out.TaskGroups[0].Services[0].Connect.SidecarService.Proxy.Expose.Paths[1] r.Equal(structs.ConsulExposePath{ Path: "/v2/health", Protocol: "grpc", LocalPathPort: 8080, ListenerPort: "v2Port", }, grpcPath) // make sure round tripping does not create duplicate expose paths out.Meta["test"] = "abc" req.Job = out r.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) r.NotZero(resp.Index) // Check for the new node in the FSM state = s1.fsm.State() ws = memdb.NewWatchSet() out, err = state.JobByID(ws, job.Namespace, job.ID) r.NoError(err) r.NotNil(out) r.Equal(resp.JobModifyIndex, out.CreateIndex) // make sure we are not re-adding what has already been added r.Len(out.TaskGroups[0].Services[0].Connect.SidecarService.Proxy.Expose.Paths, 2) } func TestJobEndpoint_Register_ConnectWithSidecarTask(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.TaskGroups[0].Networks = structs.Networks{ { Mode: "bridge", }, } job.TaskGroups[0].Tasks[0].Services = nil job.TaskGroups[0].Services = []*structs.Service{ { Name: "backend", PortLabel: "8080", Connect: &structs.ConsulConnect{ SidecarService: &structs.ConsulSidecarService{}, SidecarTask: &structs.SidecarTask{ Meta: map[string]string{ "source": "test", }, Resources: &structs.Resources{ CPU: 500, }, Config: map[string]interface{}{ "labels": map[string]string{ "foo": "bar", }, }, }, }, }, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse require.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) require.NotZero(resp.Index) // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) require.NoError(err) require.NotNil(out) require.Equal(resp.JobModifyIndex, out.CreateIndex) // Check that the sidecar task was injected require.Len(out.TaskGroups[0].Tasks, 2) sidecarTask := out.TaskGroups[0].Tasks[1] require.Equal("connect-proxy-backend", sidecarTask.Name) require.Equal("connect-proxy:backend", string(sidecarTask.Kind)) require.Equal("connect-proxy-backend", out.TaskGroups[0].Networks[0].DynamicPorts[0].Label) // Check that the correct fields were overridden from the sidecar_task stanza require.Equal("test", sidecarTask.Meta["source"]) require.Equal(500, sidecarTask.Resources.CPU) require.Equal(connectSidecarResources().MemoryMB, sidecarTask.Resources.MemoryMB) cfg := connectSidecarDriverConfig() cfg["labels"] = map[string]interface{}{ "foo": "bar", } require.Equal(cfg, sidecarTask.Config) // Check that round tripping the job doesn't change the sidecarTask out.Meta["test"] = "abc" req.Job = out require.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) require.NotZero(resp.Index) // Check for the new node in the FSM state = s1.fsm.State() ws = memdb.NewWatchSet() out, err = state.JobByID(ws, job.Namespace, job.ID) require.NoError(err) require.NotNil(out) require.Equal(resp.JobModifyIndex, out.CreateIndex) require.Len(out.TaskGroups[0].Tasks, 2) require.Exactly(sidecarTask, out.TaskGroups[0].Tasks[1]) } func TestJobEndpoint_Register_Connect_ValidatesWithoutSidecarTask(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.TaskGroups[0].Networks = structs.Networks{ { Mode: "bridge", }, } job.TaskGroups[0].Tasks[0].Services = nil job.TaskGroups[0].Services = []*structs.Service{ { Name: "backend", PortLabel: "8080", Connect: &structs.ConsulConnect{ SidecarService: nil, }, Checks: []*structs.ServiceCheck{{ Name: "exposed_no_sidecar", Type: "http", Expose: true, Path: "/health", Interval: 10 * time.Second, Timeout: 2 * time.Second, }}, }, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.Error(t, err) require.Contains(t, err.Error(), "exposed_no_sidecar requires use of sidecar_proxy") } func TestJobEndpoint_Register_ACL(t *testing.T) { t.Parallel() s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() testutil.WaitForLeader(t, s1.RPC) newVolumeJob := func(readonlyVolume bool) *structs.Job { j := mock.Job() tg := j.TaskGroups[0] tg.Volumes = map[string]*structs.VolumeRequest{ "ca-certs": { Type: structs.VolumeTypeHost, Source: "prod-ca-certs", ReadOnly: readonlyVolume, }, "csi": { Type: structs.VolumeTypeCSI, Source: "prod-db", AttachmentMode: structs.CSIVolumeAttachmentModeBlockDevice, AccessMode: structs.CSIVolumeAccessModeSingleNodeWriter, }, } tg.Tasks[0].VolumeMounts = []*structs.VolumeMount{ { Volume: "ca-certs", Destination: "/etc/ca-certificates", // Task readonly does not effect acls ReadOnly: true, }, } return j } newCSIPluginJob := func() *structs.Job { j := mock.Job() t := j.TaskGroups[0].Tasks[0] t.CSIPluginConfig = &structs.TaskCSIPluginConfig{ ID: "foo", Type: "node", } return j } submitJobPolicy := mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob, acl.NamespaceCapabilitySubmitJob}) submitJobToken := mock.CreatePolicyAndToken(t, s1.State(), 1001, "test-submit-job", submitJobPolicy) volumesPolicyReadWrite := mock.HostVolumePolicy("prod-*", "", []string{acl.HostVolumeCapabilityMountReadWrite}) volumesPolicyCSIMount := mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityCSIMountVolume}) + mock.PluginPolicy("read") submitJobWithVolumesReadWriteToken := mock.CreatePolicyAndToken(t, s1.State(), 1002, "test-submit-volumes", submitJobPolicy+ volumesPolicyReadWrite+ volumesPolicyCSIMount) volumesPolicyReadOnly := mock.HostVolumePolicy("prod-*", "", []string{acl.HostVolumeCapabilityMountReadOnly}) submitJobWithVolumesReadOnlyToken := mock.CreatePolicyAndToken(t, s1.State(), 1003, "test-submit-volumes-readonly", submitJobPolicy+ volumesPolicyReadOnly+ volumesPolicyCSIMount) pluginPolicy := mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityCSIRegisterPlugin}) pluginToken := mock.CreatePolicyAndToken(t, s1.State(), 1005, "test-csi-register-plugin", submitJobPolicy+pluginPolicy) cases := []struct { Name string Job *structs.Job Token string ErrExpected bool }{ { Name: "without a token", Job: mock.Job(), Token: "", ErrExpected: true, }, { Name: "with a token", Job: mock.Job(), Token: root.SecretID, ErrExpected: false, }, { Name: "with a token that can submit a job, but not use a required volume", Job: newVolumeJob(false), Token: submitJobToken.SecretID, ErrExpected: true, }, { Name: "with a token that can submit a job, and use all required volumes", Job: newVolumeJob(false), Token: submitJobWithVolumesReadWriteToken.SecretID, ErrExpected: false, }, { Name: "with a token that can submit a job, but only has readonly access", Job: newVolumeJob(false), Token: submitJobWithVolumesReadOnlyToken.SecretID, ErrExpected: true, }, { Name: "with a token that can submit a job, and readonly volume access is enough", Job: newVolumeJob(true), Token: submitJobWithVolumesReadOnlyToken.SecretID, ErrExpected: false, }, { Name: "with a token that can submit a job, plugin rejected", Job: newCSIPluginJob(), Token: submitJobToken.SecretID, ErrExpected: true, }, { Name: "with a token that also has csi-register-plugin, accepted", Job: newCSIPluginJob(), Token: pluginToken.SecretID, ErrExpected: false, }, } for _, tt := range cases { t.Run(tt.Name, func(t *testing.T) { codec := rpcClient(t, s1) req := &structs.JobRegisterRequest{ Job: tt.Job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: tt.Job.Namespace, }, } req.AuthToken = tt.Token // Try without a token, expect failure var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) // If we expected an error, then the job should _not_ be registered. if tt.ErrExpected { require.Error(t, err, "expected error") return } if !tt.ErrExpected { require.NoError(t, err, "unexpected error") } require.NotEqual(t, 0, resp.Index) state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, tt.Job.Namespace, tt.Job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, tt.Job.TaskGroups, out.TaskGroups) }) } } func TestJobEndpoint_Register_InvalidNamespace(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.Namespace = "foo" req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Try without a token, expect failure var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), "nonexistent namespace") { t.Fatalf("expected namespace error: %v", err) } // Check for the job in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out != nil { t.Fatalf("expected no job") } } func TestJobEndpoint_Register_Payload(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request with a job containing an invalid driver // config job := mock.Job() job.Payload = []byte{0x1} req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil { t.Fatalf("expected a validation error") } if !strings.Contains(err.Error(), "payload") { t.Fatalf("expected a payload error but got: %v", err) } } func TestJobEndpoint_Register_Existing(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Update the job definition job2 := mock.Job() job2.Priority = 100 job2.ID = job.ID req.Job = job2 // Attempt update if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.ModifyIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } if out.Priority != 100 { t.Fatalf("expected update") } if out.Version != 1 { t.Fatalf("expected update") } // Lookup the evaluation eval, err := state.EvalByID(ws, resp.EvalID) if err != nil { t.Fatalf("err: %v", err) } if eval == nil { t.Fatalf("expected eval") } if eval.CreateIndex != resp.EvalCreateIndex { t.Fatalf("index mis-match") } if eval.Priority != job2.Priority { t.Fatalf("bad: %#v", eval) } if eval.Type != job2.Type { t.Fatalf("bad: %#v", eval) } if eval.TriggeredBy != structs.EvalTriggerJobRegister { t.Fatalf("bad: %#v", eval) } if eval.JobID != job2.ID { t.Fatalf("bad: %#v", eval) } if eval.JobModifyIndex != resp.JobModifyIndex { t.Fatalf("bad: %#v", eval) } if eval.Status != structs.EvalStatusPending { t.Fatalf("bad: %#v", eval) } if eval.CreateTime == 0 { t.Fatalf("eval CreateTime is unset: %#v", eval) } if eval.ModifyTime == 0 { t.Fatalf("eval ModifyTime is unset: %#v", eval) } if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Check to ensure the job version didn't get bumped because we submitted // the same job state = s1.fsm.State() ws = memdb.NewWatchSet() out, err = state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.Version != 1 { t.Fatalf("expected no update; got %v; diff %v", out.Version, pretty.Diff(job2, out)) } } func TestJobEndpoint_Register_Periodic(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request for a periodic job. job := mock.PeriodicJob() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.JobModifyIndex == 0 { t.Fatalf("bad index: %d", resp.Index) } // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } serviceName := out.TaskGroups[0].Tasks[0].Services[0].Name expectedServiceName := "web-frontend" if serviceName != expectedServiceName { t.Fatalf("Expected Service Name: %s, Actual: %s", expectedServiceName, serviceName) } if resp.EvalID != "" { t.Fatalf("Register created an eval for a periodic job") } } func TestJobEndpoint_Register_ParameterizedJob(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request for a parameterized job. job := mock.BatchJob() job.ParameterizedJob = &structs.ParameterizedJobConfig{} req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.JobModifyIndex == 0 { t.Fatalf("bad index: %d", resp.Index) } // Check for the job in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } if resp.EvalID != "" { t.Fatalf("Register created an eval for a parameterized job") } } func TestJobEndpoint_Register_Dispatched(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request with a job with 'Dispatch' set to true job := mock.Job() job.Dispatched = true job.ParameterizedJob = &structs.ParameterizedJobConfig{} req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.Error(err) require.Contains(err.Error(), "job can't be submitted with 'Dispatched'") } func TestJobEndpoint_Register_EnforceIndex(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request and enforcing an incorrect index job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, EnforceIndex: true, JobModifyIndex: 100, // Not registered yet so not possible WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), RegisterEnforceIndexErrPrefix) { t.Fatalf("expected enforcement error") } // Create the register request and enforcing it is new req = &structs.JobRegisterRequest{ Job: job, EnforceIndex: true, JobModifyIndex: 0, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } curIndex := resp.JobModifyIndex // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } // Reregister request and enforcing it be a new job req = &structs.JobRegisterRequest{ Job: job, EnforceIndex: true, JobModifyIndex: 0, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), RegisterEnforceIndexErrPrefix) { t.Fatalf("expected enforcement error") } // Reregister request and enforcing it be at an incorrect index req = &structs.JobRegisterRequest{ Job: job, EnforceIndex: true, JobModifyIndex: curIndex - 1, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), RegisterEnforceIndexErrPrefix) { t.Fatalf("expected enforcement error") } // Reregister request and enforcing it be at the correct index job.Priority = job.Priority + 1 req = &structs.JobRegisterRequest{ Job: job, EnforceIndex: true, JobModifyIndex: curIndex, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } out, err = state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.Priority != job.Priority { t.Fatalf("priority mis-match") } } // TestJobEndpoint_Register_Vault_Disabled asserts that submitting a job that // uses Vault when Vault is *disabled* results in an error. func TestJobEndpoint_Register_Vault_Disabled(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue f := false c.VaultConfig.Enabled = &f }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request with a job asking for a vault policy job := mock.Job() job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{"foo"}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), "Vault not enabled") { t.Fatalf("expected Vault not enabled error: %v", err) } } // TestJobEndpoint_Register_Vault_AllowUnauthenticated asserts submitting a job // with a Vault policy but without a Vault token is *succeeds* if // allow_unauthenticated=true. func TestJobEndpoint_Register_Vault_AllowUnauthenticated(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault and allow authenticated tr := true s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &tr // Replace the Vault Client on the server s1.vault = &TestVaultClient{} // Create the register request with a job asking for a vault policy job := mock.Job() job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{"foo"}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err != nil { t.Fatalf("bad: %v", err) } // Check for the job in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } } // TestJobEndpoint_Register_Vault_OverrideConstraint asserts that job // submitters can specify their own Vault constraint to override the // automatically injected one. func TestJobEndpoint_Register_Vault_OverrideConstraint(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault and allow authenticated tr := true s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &tr // Replace the Vault Client on the server s1.vault = &TestVaultClient{} // Create the register request with a job asking for a vault policy job := mock.Job() job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{"foo"}, ChangeMode: structs.VaultChangeModeRestart, } job.TaskGroups[0].Tasks[0].Constraints = []*structs.Constraint{ { LTarget: "${attr.vault.version}", Operand: "is_set", }, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) // Check for the job in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp.JobModifyIndex, out.CreateIndex) // Assert constraint was not overridden by the server outConstraints := out.TaskGroups[0].Tasks[0].Constraints require.Len(t, outConstraints, 1) require.True(t, job.TaskGroups[0].Tasks[0].Constraints[0].Equals(outConstraints[0])) } func TestJobEndpoint_Register_Vault_NoToken(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault tr, f := true, false s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &f // Replace the Vault Client on the server s1.vault = &TestVaultClient{} // Create the register request with a job asking for a vault policy but // don't send a Vault token job := mock.Job() job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{"foo"}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), "missing Vault Token") { t.Fatalf("expected Vault not enabled error: %v", err) } } func TestJobEndpoint_Register_Vault_Policies(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault tr, f := true, false s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &f // Replace the Vault Client on the server tvc := &TestVaultClient{} s1.vault = tvc // Add three tokens: one that allows the requesting policy, one that does // not and one that returns an error policy := "foo" badToken := uuid.Generate() badPolicies := []string{"a", "b", "c"} tvc.SetLookupTokenAllowedPolicies(badToken, badPolicies) goodToken := uuid.Generate() goodPolicies := []string{"foo", "bar", "baz"} tvc.SetLookupTokenAllowedPolicies(goodToken, goodPolicies) rootToken := uuid.Generate() rootPolicies := []string{"root"} tvc.SetLookupTokenAllowedPolicies(rootToken, rootPolicies) errToken := uuid.Generate() expectedErr := fmt.Errorf("return errors from vault") tvc.SetLookupTokenError(errToken, expectedErr) // Create the register request with a job asking for a vault policy but // send the bad Vault token job := mock.Job() job.VaultToken = badToken job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{policy}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), "doesn't allow access to the following policies: "+policy) { t.Fatalf("expected permission denied error: %v", err) } // Use the err token job.VaultToken = errToken err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if err == nil || !strings.Contains(err.Error(), expectedErr.Error()) { t.Fatalf("expected permission denied error: %v", err) } // Use the good token job.VaultToken = goodToken // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("bad: %v", err) } // Check for the job in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } if out.VaultToken != "" { t.Fatalf("vault token not cleared") } // Check that an implicit constraint was created constraints := out.TaskGroups[0].Constraints if l := len(constraints); l != 1 { t.Fatalf("Unexpected number of tests: %v", l) } if !constraints[0].Equal(vaultConstraint) { t.Fatalf("bad constraint; got %#v; want %#v", constraints[0], vaultConstraint) } // Create the register request with another job asking for a vault policy but // send the root Vault token job2 := mock.Job() job2.VaultToken = rootToken job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{policy}, ChangeMode: structs.VaultChangeModeRestart, } req = &structs.JobRegisterRequest{ Job: job2, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("bad: %v", err) } // Check for the job in the FSM out, err = state.JobByID(ws, job2.Namespace, job2.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } if out.VaultToken != "" { t.Fatalf("vault token not cleared") } } func TestJobEndpoint_Register_Vault_MultiNamespaces(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault tr, f := true, false s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &f // Replace the Vault Client on the server tvc := &TestVaultClient{} s1.vault = tvc goodToken := uuid.Generate() goodPolicies := []string{"foo", "bar", "baz"} tvc.SetLookupTokenAllowedPolicies(goodToken, goodPolicies) // Create the register request with a job asking for a vault policy but // don't send a Vault token job := mock.Job() job.VaultToken = goodToken job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Namespace: "ns1", Policies: []string{"foo"}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) // OSS or Ent check if err != nil && s1.EnterpriseState.Features() == 0 { // errors.Is cannot be used because the RPC call break error wrapping. require.Contains(t, err.Error(), ErrMultipleNamespaces.Error()) } else { require.NoError(t, err) } } // TestJobEndpoint_Register_SemverConstraint asserts that semver ordering is // used when evaluating semver constraints. func TestJobEndpoint_Register_SemverConstraint(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.State() // Create a job with a semver constraint job := mock.Job() job.Constraints = []*structs.Constraint{ { LTarget: "${attr.vault.version}", RTarget: ">= 0.6.1", Operand: structs.ConstraintSemver, }, } job.TaskGroups[0].Count = 1 // Insert 2 Nodes, 1 matching the constraint, 1 not node1 := mock.Node() node1.Attributes["vault.version"] = "1.3.0-beta1+ent" node1.ComputeClass() require.NoError(t, state.UpsertNode(structs.MsgTypeTestSetup, 1, node1)) node2 := mock.Node() delete(node2.Attributes, "vault.version") node2.ComputeClass() require.NoError(t, state.UpsertNode(structs.MsgTypeTestSetup, 2, node2)) // Create the register request req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse require.NoError(t, msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp)) require.NotZero(t, resp.Index) // Wait for placements allocReq := &structs.JobSpecificRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: structs.DefaultNamespace, }, } testutil.WaitForResult(func() (bool, error) { resp := structs.JobAllocationsResponse{} err := msgpackrpc.CallWithCodec(codec, "Job.Allocations", allocReq, &resp) if err != nil { return false, err } if n := len(resp.Allocations); n != 1 { return false, fmt.Errorf("expected 1 alloc, found %d", n) } alloc := resp.Allocations[0] if alloc.NodeID != node1.ID { return false, fmt.Errorf("expected alloc to be one node=%q but found node=%q", node1.ID, alloc.NodeID) } return true, nil }, func(waitErr error) { evals, err := state.EvalsByJob(nil, structs.DefaultNamespace, job.ID) require.NoError(t, err) for i, e := range evals { t.Logf("%d Eval: %s", i, pretty.Sprint(e)) } require.NoError(t, waitErr) }) } // TestJobEndpoint_Register_EvalCreation_Modern asserts that job register creates an eval // atomically with the registration func TestJobEndpoint_Register_EvalCreation_Modern(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request t.Run("job registration always create evals", func(t *testing.T) { job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } //// initial registration should create the job and a new eval var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) require.NotZero(t, resp.Index) require.NotEmpty(t, resp.EvalID) // Check for the job in the FSM state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp.JobModifyIndex, out.CreateIndex) // Lookup the evaluation eval, err := state.EvalByID(nil, resp.EvalID) require.NoError(t, err) require.NotNil(t, eval) require.Equal(t, resp.EvalCreateIndex, eval.CreateIndex) require.Nil(t, evalUpdateFromRaft(t, s1, eval.ID)) //// re-registration should create a new eval, but leave the job untouched var resp2 structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp2) require.NoError(t, err) require.NotZero(t, resp2.Index) require.NotEmpty(t, resp2.EvalID) require.NotEqual(t, resp.EvalID, resp2.EvalID) // Check for the job in the FSM state = s1.fsm.State() out, err = state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp2.JobModifyIndex, out.CreateIndex) require.Equal(t, out.CreateIndex, out.JobModifyIndex) // Lookup the evaluation eval, err = state.EvalByID(nil, resp2.EvalID) require.NoError(t, err) require.NotNil(t, eval) require.Equal(t, resp2.EvalCreateIndex, eval.CreateIndex) raftEval := evalUpdateFromRaft(t, s1, eval.ID) require.Equal(t, raftEval, eval) //// an update should update the job and create a new eval req.Job.TaskGroups[0].Name += "a" var resp3 structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp3) require.NoError(t, err) require.NotZero(t, resp3.Index) require.NotEmpty(t, resp3.EvalID) require.NotEqual(t, resp.EvalID, resp3.EvalID) // Check for the job in the FSM state = s1.fsm.State() out, err = state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp3.JobModifyIndex, out.JobModifyIndex) // Lookup the evaluation eval, err = state.EvalByID(nil, resp3.EvalID) require.NoError(t, err) require.NotNil(t, eval) require.Equal(t, resp3.EvalCreateIndex, eval.CreateIndex) require.Nil(t, evalUpdateFromRaft(t, s1, eval.ID)) }) // Registering a parameterized job shouldn't create an eval t.Run("periodic jobs shouldn't create an eval", func(t *testing.T) { job := mock.PeriodicJob() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) require.NotZero(t, resp.Index) require.Empty(t, resp.EvalID) // Check for the job in the FSM state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp.JobModifyIndex, out.CreateIndex) }) } // TestJobEndpoint_Register_EvalCreation_Legacy asserts that job register creates an eval // atomically with the registration, but handle legacy clients by adding a new eval update func TestJobEndpoint_Register_EvalCreation_Legacy(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.BootstrapExpect = 2 c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() s2, cleanupS2 := TestServer(t, func(c *Config) { c.BootstrapExpect = 2 c.NumSchedulers = 0 // Prevent automatic dequeue // simulate presense of a server that doesn't handle // new registration eval c.Build = "0.12.0" }) defer cleanupS2() TestJoin(t, s1, s2) testutil.WaitForLeader(t, s1.RPC) testutil.WaitForLeader(t, s2.RPC) // keep s1 as the leader if leader, _ := s1.getLeader(); !leader { s1, s2 = s2, s1 } codec := rpcClient(t, s1) // Create the register request t.Run("job registration always create evals", func(t *testing.T) { job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } //// initial registration should create the job and a new eval var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) require.NotZero(t, resp.Index) require.NotEmpty(t, resp.EvalID) // Check for the job in the FSM state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp.JobModifyIndex, out.CreateIndex) // Lookup the evaluation eval, err := state.EvalByID(nil, resp.EvalID) require.NoError(t, err) require.NotNil(t, eval) require.Equal(t, resp.EvalCreateIndex, eval.CreateIndex) raftEval := evalUpdateFromRaft(t, s1, eval.ID) require.Equal(t, eval, raftEval) //// re-registration should create a new eval, but leave the job untouched var resp2 structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp2) require.NoError(t, err) require.NotZero(t, resp2.Index) require.NotEmpty(t, resp2.EvalID) require.NotEqual(t, resp.EvalID, resp2.EvalID) // Check for the job in the FSM state = s1.fsm.State() out, err = state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp2.JobModifyIndex, out.CreateIndex) require.Equal(t, out.CreateIndex, out.JobModifyIndex) // Lookup the evaluation eval, err = state.EvalByID(nil, resp2.EvalID) require.NoError(t, err) require.NotNil(t, eval) require.Equal(t, resp2.EvalCreateIndex, eval.CreateIndex) // this raft eval is the one found above raftEval = evalUpdateFromRaft(t, s1, eval.ID) require.Equal(t, eval, raftEval) //// an update should update the job and create a new eval req.Job.TaskGroups[0].Name += "a" var resp3 structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp3) require.NoError(t, err) require.NotZero(t, resp3.Index) require.NotEmpty(t, resp3.EvalID) require.NotEqual(t, resp.EvalID, resp3.EvalID) // Check for the job in the FSM state = s1.fsm.State() out, err = state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp3.JobModifyIndex, out.JobModifyIndex) // Lookup the evaluation eval, err = state.EvalByID(nil, resp3.EvalID) require.NoError(t, err) require.NotNil(t, eval) require.Equal(t, resp3.EvalCreateIndex, eval.CreateIndex) raftEval = evalUpdateFromRaft(t, s1, eval.ID) require.Equal(t, eval, raftEval) }) // Registering a parameterized job shouldn't create an eval t.Run("periodic jobs shouldn't create an eval", func(t *testing.T) { job := mock.PeriodicJob() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) require.NotZero(t, resp.Index) require.Empty(t, resp.EvalID) // Check for the job in the FSM state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, resp.JobModifyIndex, out.CreateIndex) }) } func TestJobEndpoint_Register_ValidateMemoryMax(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request submitNewJob := func() *structs.JobRegisterResponse { job := mock.Job() job.TaskGroups[0].Tasks[0].Resources.MemoryMaxMB = 2000 req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) return &resp } // try default case: Memory oversubscription is disabled resp := submitNewJob() require.Contains(t, resp.Warnings, "Memory oversubscription is not enabled") // enable now and try again s1.State().SchedulerSetConfig(100, &structs.SchedulerConfiguration{ MemoryOversubscriptionEnabled: true, }) resp = submitNewJob() require.Empty(t, resp.Warnings) } // evalUpdateFromRaft searches the raft logs for the eval update pertaining to the eval func evalUpdateFromRaft(t *testing.T, s *Server, evalID string) *structs.Evaluation { var store raft.LogStore = s.raftInmem if store == nil { store = s.raftStore } require.NotNil(t, store) li, _ := store.LastIndex() for i, _ := store.FirstIndex(); i <= li; i++ { var log raft.Log err := store.GetLog(i, &log) require.NoError(t, err) if log.Type != raft.LogCommand { continue } if structs.MessageType(log.Data[0]) != structs.EvalUpdateRequestType { continue } var req structs.EvalUpdateRequest structs.Decode(log.Data[1:], &req) require.NoError(t, err) for _, eval := range req.Evals { if eval.ID == evalID { eval.CreateIndex = i eval.ModifyIndex = i return eval } } } return nil } func TestJobEndpoint_Register_ACL_Namespace(t *testing.T) { t.Parallel() s1, _, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Policy with read on default namespace and write on non default policy := &structs.ACLPolicy{ Name: fmt.Sprintf("policy-%s", uuid.Generate()), Description: "Super cool policy!", Rules: ` namespace "default" { policy = "read" } namespace "test" { policy = "write" } node { policy = "read" } agent { policy = "read" } `, CreateIndex: 10, ModifyIndex: 20, } policy.SetHash() assert := assert.New(t) // Upsert policy and token token := mock.ACLToken() token.Policies = []string{policy.Name} err := s1.State().UpsertACLPolicies(structs.MsgTypeTestSetup, 100, []*structs.ACLPolicy{policy}) assert.Nil(err) err = s1.State().UpsertACLTokens(structs.MsgTypeTestSetup, 110, []*structs.ACLToken{token}) assert.Nil(err) // Upsert namespace ns := mock.Namespace() ns.Name = "test" err = s1.fsm.State().UpsertNamespaces(1000, []*structs.Namespace{ns}) assert.Nil(err) // Create the register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{Region: "global"}, } req.AuthToken = token.SecretID // Use token without write access to default namespace, expect failure var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) assert.NotNil(err, "expected permission denied") req.Namespace = "test" job.Namespace = "test" // Use token with write access to default namespace, expect success err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) assert.Nil(err, "unexpected err: %v", err) assert.NotEqual(resp.Index, 0, "bad index: %d", resp.Index) // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) assert.Nil(err) assert.NotNil(out, "expected job") } func TestJobRegister_ACL_RejectedBySchedulerConfig(t *testing.T) { t.Parallel() s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) submitJobToken := mock.CreatePolicyAndToken(t, s1.State(), 1001, "test-valid-write", mock.NamespacePolicy(structs.DefaultNamespace, "write", nil)). SecretID cases := []struct { name string token string rejectEnabled bool errExpected string }{ { name: "reject disabled, with a submit token", token: submitJobToken, rejectEnabled: false, }, { name: "reject enabled, with a submit token", token: submitJobToken, rejectEnabled: true, errExpected: structs.ErrJobRegistrationDisabled.Error(), }, { name: "reject enabled, without a token", token: "", rejectEnabled: true, errExpected: structs.ErrPermissionDenied.Error(), }, { name: "reject enabled, with a management token", token: root.SecretID, rejectEnabled: true, }, } for _, tc := range cases { t.Run(tc.name, func(t *testing.T) { job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } req.AuthToken = tc.token cfgReq := &structs.SchedulerSetConfigRequest{ Config: structs.SchedulerConfiguration{ RejectJobRegistration: tc.rejectEnabled, }, WriteRequest: structs.WriteRequest{ Region: "global", }, } cfgReq.AuthToken = root.SecretID err := msgpackrpc.CallWithCodec(codec, "Operator.SchedulerSetConfiguration", cfgReq, &structs.SchedulerSetConfigurationResponse{}, ) require.NoError(t, err) var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) if tc.errExpected != "" { require.Error(t, err, "expected error") require.EqualError(t, err, tc.errExpected) state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) require.NoError(t, err) require.Nil(t, out) } else { require.NoError(t, err, "unexpected error") require.NotEqual(t, 0, resp.Index) state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) require.NoError(t, err) require.NotNil(t, out) require.Equal(t, job.TaskGroups, out.TaskGroups) } }) } } func TestJobEndpoint_Revert(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the initial register request job := mock.Job() job.Priority = 100 req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Reregister again to get another version job2 := job.Copy() job2.Priority = 1 req = &structs.JobRegisterRequest{ Job: job2, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Create revert request and enforcing it be at an incorrect version revertReq := &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 0, EnforcePriorVersion: helper.Uint64ToPtr(10), WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response err := msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp) if err == nil || !strings.Contains(err.Error(), "enforcing version 10") { t.Fatalf("expected enforcement error") } // Create revert request and enforcing it be at the current version revertReq = &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 1, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response err = msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp) if err == nil || !strings.Contains(err.Error(), "current version") { t.Fatalf("expected current version err: %v", err) } // Create revert request and enforcing it be at version 1 revertReq = &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 0, EnforcePriorVersion: helper.Uint64ToPtr(1), WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } if resp.EvalID == "" || resp.EvalCreateIndex == 0 { t.Fatalf("bad created eval: %+v", resp) } if resp.JobModifyIndex == 0 { t.Fatalf("bad job modify index: %d", resp.JobModifyIndex) } // Create revert request and don't enforce. We are at version 2 but it is // the same as version 0 revertReq = &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 0, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } if resp.EvalID == "" || resp.EvalCreateIndex == 0 { t.Fatalf("bad created eval: %+v", resp) } if resp.JobModifyIndex == 0 { t.Fatalf("bad job modify index: %d", resp.JobModifyIndex) } // Check that the job is at the correct version and that the eval was // created state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.Priority != job.Priority { t.Fatalf("priority mis-match") } if out.Version != 2 { t.Fatalf("got version %d; want %d", out.Version, 2) } eout, err := state.EvalByID(ws, resp.EvalID) if err != nil { t.Fatalf("err: %v", err) } if eout == nil { t.Fatalf("expected eval") } if eout.JobID != job.ID { t.Fatalf("job id mis-match") } versions, err := state.JobVersionsByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if len(versions) != 3 { t.Fatalf("got %d versions; want %d", len(versions), 3) } } func TestJobEndpoint_Revert_Vault_NoToken(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault tr, f := true, false s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &f // Replace the Vault Client on the server tvc := &TestVaultClient{} s1.vault = tvc // Add three tokens: one that allows the requesting policy, one that does // not and one that returns an error policy := "foo" goodToken := uuid.Generate() goodPolicies := []string{"foo", "bar", "baz"} tvc.SetLookupTokenAllowedPolicies(goodToken, goodPolicies) // Create the initial register request job := mock.Job() job.VaultToken = goodToken job.Priority = 100 job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{policy}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } // Reregister again to get another version job2 := job.Copy() job2.Priority = 1 req = &structs.JobRegisterRequest{ Job: job2, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } revertReq := &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 1, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response err := msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp) if err == nil || !strings.Contains(err.Error(), "current version") { t.Fatalf("expected current version err: %v", err) } // Create revert request and enforcing it be at version 1 revertReq = &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 0, EnforcePriorVersion: helper.Uint64ToPtr(1), WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response err = msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp) if err == nil || !strings.Contains(err.Error(), "missing Vault Token") { t.Fatalf("expected Vault not enabled error: %v", err) } } // TestJobEndpoint_Revert_Vault_Policies asserts that job revert uses the // revert request's Vault token when authorizing policies. func TestJobEndpoint_Revert_Vault_Policies(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault tr, f := true, false s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &f // Replace the Vault Client on the server tvc := &TestVaultClient{} s1.vault = tvc // Add three tokens: one that allows the requesting policy, one that does // not and one that returns an error policy := "foo" badToken := uuid.Generate() badPolicies := []string{"a", "b", "c"} tvc.SetLookupTokenAllowedPolicies(badToken, badPolicies) registerGoodToken := uuid.Generate() goodPolicies := []string{"foo", "bar", "baz"} tvc.SetLookupTokenAllowedPolicies(registerGoodToken, goodPolicies) revertGoodToken := uuid.Generate() revertGoodPolicies := []string{"foo", "bar_revert", "baz_revert"} tvc.SetLookupTokenAllowedPolicies(revertGoodToken, revertGoodPolicies) rootToken := uuid.Generate() rootPolicies := []string{"root"} tvc.SetLookupTokenAllowedPolicies(rootToken, rootPolicies) errToken := uuid.Generate() expectedErr := fmt.Errorf("return errors from vault") tvc.SetLookupTokenError(errToken, expectedErr) // Create the initial register request job := mock.Job() job.VaultToken = registerGoodToken job.Priority = 100 job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{policy}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } // Reregister again to get another version job2 := job.Copy() job2.Priority = 1 req = &structs.JobRegisterRequest{ Job: job2, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } // Create the revert request with the bad Vault token revertReq := &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 0, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, VaultToken: badToken, } // Fetch the response err := msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp) if err == nil || !strings.Contains(err.Error(), "doesn't allow access to the following policies: "+policy) { t.Fatalf("expected permission denied error: %v", err) } // Use the err token revertReq.VaultToken = errToken err = msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp) if err == nil || !strings.Contains(err.Error(), expectedErr.Error()) { t.Fatalf("expected permission denied error: %v", err) } // Use a good token revertReq.VaultToken = revertGoodToken if err := msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp); err != nil { t.Fatalf("bad: %v", err) } } func TestJobEndpoint_Revert_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) state := s1.fsm.State() testutil.WaitForLeader(t, s1.RPC) // Create the jobs job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 300, job) require.Nil(err) job2 := job.Copy() job2.Priority = 1 err = state.UpsertJob(structs.MsgTypeTestSetup, 400, job2) require.Nil(err) // Create revert request and enforcing it be at the current version revertReq := &structs.JobRevertRequest{ JobID: job.ID, JobVersion: 0, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Attempt to fetch the response without a valid token var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1001, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) revertReq.AuthToken = invalidToken.SecretID var invalidResp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Fetch the response with a valid management token revertReq.AuthToken = root.SecretID var validResp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &validResp) require.Nil(err) // Try with a valid non-management token validToken := mock.CreatePolicyAndToken(t, state, 1003, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilitySubmitJob})) revertReq.AuthToken = validToken.SecretID var validResp2 structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Revert", revertReq, &validResp2) require.Nil(err) } func TestJobEndpoint_Stable(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the initial register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Create stability request stableReq := &structs.JobStabilityRequest{ JobID: job.ID, JobVersion: 0, Stable: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var stableResp structs.JobStabilityResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Stable", stableReq, &stableResp); err != nil { t.Fatalf("err: %v", err) } if stableResp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Check that the job is marked stable state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if !out.Stable { t.Fatalf("Job is not marked stable") } } func TestJobEndpoint_Stable_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) state := s1.fsm.State() testutil.WaitForLeader(t, s1.RPC) // Register the job job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) // Create stability request stableReq := &structs.JobStabilityRequest{ JobID: job.ID, JobVersion: 0, Stable: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Attempt to fetch the token without a token var stableResp structs.JobStabilityResponse err = msgpackrpc.CallWithCodec(codec, "Job.Stable", stableReq, &stableResp) require.NotNil(err) require.Contains("Permission denied", err.Error()) // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) stableReq.AuthToken = invalidToken.SecretID var invalidStableResp structs.JobStabilityResponse err = msgpackrpc.CallWithCodec(codec, "Job.Stable", stableReq, &invalidStableResp) require.NotNil(err) require.Contains("Permission denied", err.Error()) // Attempt to fetch with a management token stableReq.AuthToken = root.SecretID var validStableResp structs.JobStabilityResponse err = msgpackrpc.CallWithCodec(codec, "Job.Stable", stableReq, &validStableResp) require.Nil(err) // Attempt to fetch with a valid token validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilitySubmitJob})) stableReq.AuthToken = validToken.SecretID var validStableResp2 structs.JobStabilityResponse err = msgpackrpc.CallWithCodec(codec, "Job.Stable", stableReq, &validStableResp2) require.Nil(err) // Check that the job is marked stable ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) require.Nil(err) require.NotNil(job) require.Equal(true, out.Stable) } func TestJobEndpoint_Evaluate(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Force a re-evaluation reEval := &structs.JobEvaluateRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Lookup the evaluation state := s1.fsm.State() ws := memdb.NewWatchSet() eval, err := state.EvalByID(ws, resp.EvalID) if err != nil { t.Fatalf("err: %v", err) } if eval == nil { t.Fatalf("expected eval") } if eval.CreateIndex != resp.EvalCreateIndex { t.Fatalf("index mis-match") } if eval.Priority != job.Priority { t.Fatalf("bad: %#v", eval) } if eval.Type != job.Type { t.Fatalf("bad: %#v", eval) } if eval.TriggeredBy != structs.EvalTriggerJobRegister { t.Fatalf("bad: %#v", eval) } if eval.JobID != job.ID { t.Fatalf("bad: %#v", eval) } if eval.JobModifyIndex != resp.JobModifyIndex { t.Fatalf("bad: %#v", eval) } if eval.Status != structs.EvalStatusPending { t.Fatalf("bad: %#v", eval) } if eval.CreateTime == 0 { t.Fatalf("eval CreateTime is unset: %#v", eval) } if eval.ModifyTime == 0 { t.Fatalf("eval ModifyTime is unset: %#v", eval) } } func TestJobEndpoint_ForceRescheduleEvaluate(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.Nil(err) require.NotEqual(0, resp.Index) state := s1.fsm.State() job, err = state.JobByID(nil, structs.DefaultNamespace, job.ID) require.Nil(err) // Create a failed alloc alloc := mock.Alloc() alloc.Job = job alloc.JobID = job.ID alloc.TaskGroup = job.TaskGroups[0].Name alloc.Namespace = job.Namespace alloc.ClientStatus = structs.AllocClientStatusFailed err = s1.State().UpsertAllocs(structs.MsgTypeTestSetup, resp.Index+1, []*structs.Allocation{alloc}) require.Nil(err) // Force a re-evaluation reEval := &structs.JobEvaluateRequest{ JobID: job.ID, EvalOptions: structs.EvalOptions{ForceReschedule: true}, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response err = msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &resp) require.Nil(err) require.NotEqual(0, resp.Index) // Lookup the evaluation ws := memdb.NewWatchSet() eval, err := state.EvalByID(ws, resp.EvalID) require.Nil(err) require.NotNil(eval) require.Equal(eval.CreateIndex, resp.EvalCreateIndex) require.Equal(eval.Priority, job.Priority) require.Equal(eval.Type, job.Type) require.Equal(eval.TriggeredBy, structs.EvalTriggerJobRegister) require.Equal(eval.JobID, job.ID) require.Equal(eval.JobModifyIndex, resp.JobModifyIndex) require.Equal(eval.Status, structs.EvalStatusPending) require.NotZero(eval.CreateTime) require.NotZero(eval.ModifyTime) // Lookup the alloc, verify DesiredTransition ForceReschedule alloc, err = state.AllocByID(ws, alloc.ID) require.NotNil(alloc) require.Nil(err) require.True(*alloc.DesiredTransition.ForceReschedule) } func TestJobEndpoint_Evaluate_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create the job job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 300, job) require.Nil(err) // Force a re-evaluation reEval := &structs.JobEvaluateRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Attempt to fetch the response without a token var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) reEval.AuthToken = invalidToken.SecretID var invalidResp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Fetch the response with a valid management token reEval.AuthToken = root.SecretID var validResp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &validResp) require.Nil(err) // Fetch the response with a valid token validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) reEval.AuthToken = validToken.SecretID var validResp2 structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &validResp2) require.Nil(err) // Lookup the evaluation ws := memdb.NewWatchSet() eval, err := state.EvalByID(ws, validResp2.EvalID) require.Nil(err) require.NotNil(eval) require.Equal(eval.CreateIndex, validResp2.EvalCreateIndex) require.Equal(eval.Priority, job.Priority) require.Equal(eval.Type, job.Type) require.Equal(eval.TriggeredBy, structs.EvalTriggerJobRegister) require.Equal(eval.JobID, job.ID) require.Equal(eval.JobModifyIndex, validResp2.JobModifyIndex) require.Equal(eval.Status, structs.EvalStatusPending) require.NotZero(eval.CreateTime) require.NotZero(eval.ModifyTime) } func TestJobEndpoint_Evaluate_Periodic(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.PeriodicJob() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.JobModifyIndex == 0 { t.Fatalf("bad index: %d", resp.Index) } // Force a re-evaluation reEval := &structs.JobEvaluateRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &resp); err == nil { t.Fatal("expect an err") } } func TestJobEndpoint_Evaluate_ParameterizedJob(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.BatchJob() job.ParameterizedJob = &structs.ParameterizedJobConfig{} req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.JobModifyIndex == 0 { t.Fatalf("bad index: %d", resp.Index) } // Force a re-evaluation reEval := &structs.JobEvaluateRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response if err := msgpackrpc.CallWithCodec(codec, "Job.Evaluate", reEval, &resp); err == nil { t.Fatal("expect an err") } } func TestJobEndpoint_Deregister(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register requests job := mock.Job() reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp)) // Deregister but don't purge dereg := &structs.JobDeregisterRequest{ JobID: job.ID, Purge: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2)) require.NotZero(resp2.Index) // Check for the job in the FSM state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.Nil(err) require.NotNil(out) require.True(out.Stop) // Lookup the evaluation eval, err := state.EvalByID(nil, resp2.EvalID) require.Nil(err) require.NotNil(eval) require.EqualValues(resp2.EvalCreateIndex, eval.CreateIndex) require.Equal(job.Priority, eval.Priority) require.Equal(job.Type, eval.Type) require.Equal(structs.EvalTriggerJobDeregister, eval.TriggeredBy) require.Equal(job.ID, eval.JobID) require.Equal(structs.EvalStatusPending, eval.Status) require.NotZero(eval.CreateTime) require.NotZero(eval.ModifyTime) // Deregister and purge dereg2 := &structs.JobDeregisterRequest{ JobID: job.ID, Purge: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp3 structs.JobDeregisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg2, &resp3)) require.NotZero(resp3.Index) // Check for the job in the FSM out, err = state.JobByID(nil, job.Namespace, job.ID) require.Nil(err) require.Nil(out) // Lookup the evaluation eval, err = state.EvalByID(nil, resp3.EvalID) require.Nil(err) require.NotNil(eval) require.EqualValues(resp3.EvalCreateIndex, eval.CreateIndex) require.Equal(job.Priority, eval.Priority) require.Equal(job.Type, eval.Type) require.Equal(structs.EvalTriggerJobDeregister, eval.TriggeredBy) require.Equal(job.ID, eval.JobID) require.Equal(structs.EvalStatusPending, eval.Status) require.NotZero(eval.CreateTime) require.NotZero(eval.ModifyTime) } func TestJobEndpoint_Deregister_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create and register a job job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 100, job) require.Nil(err) // Deregister and purge req := &structs.JobDeregisterRequest{ JobID: job.ID, Purge: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Expect failure for request without a token var resp structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", req, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) req.AuthToken = invalidToken.SecretID var invalidResp structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", req, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Expect success with a valid management token req.AuthToken = root.SecretID var validResp structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", req, &validResp) require.Nil(err) require.NotEqual(validResp.Index, 0) // Expect success with a valid token validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilitySubmitJob})) req.AuthToken = validToken.SecretID // Check for the job in the FSM out, err := state.JobByID(nil, job.Namespace, job.ID) require.Nil(err) require.Nil(out) // Lookup the evaluation eval, err := state.EvalByID(nil, validResp.EvalID) require.Nil(err) require.NotNil(eval, nil) require.Equal(eval.CreateIndex, validResp.EvalCreateIndex) require.Equal(eval.Priority, structs.JobDefaultPriority) require.Equal(eval.Type, structs.JobTypeService) require.Equal(eval.TriggeredBy, structs.EvalTriggerJobDeregister) require.Equal(eval.JobID, job.ID) require.Equal(eval.JobModifyIndex, validResp.JobModifyIndex) require.Equal(eval.Status, structs.EvalStatusPending) require.NotZero(eval.CreateTime) require.NotZero(eval.ModifyTime) // Deregistration is not idempotent, produces a new eval after the job is // deregistered. TODO(langmartin) make it idempotent. var validResp2 structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", req, &validResp2) require.NoError(err) require.NotEqual("", validResp2.EvalID) require.NotEqual(validResp.EvalID, validResp2.EvalID) } func TestJobEndpoint_Deregister_Nonexistent(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Deregister jobID := "foo" dereg := &structs.JobDeregisterRequest{ JobID: jobID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: structs.DefaultNamespace, }, } var resp2 structs.JobDeregisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.JobModifyIndex == 0 { t.Fatalf("bad index: %d", resp2.Index) } // Lookup the evaluation state := s1.fsm.State() ws := memdb.NewWatchSet() eval, err := state.EvalByID(ws, resp2.EvalID) if err != nil { t.Fatalf("err: %v", err) } if eval == nil { t.Fatalf("expected eval") } if eval.CreateIndex != resp2.EvalCreateIndex { t.Fatalf("index mis-match") } if eval.Priority != structs.JobDefaultPriority { t.Fatalf("bad: %#v", eval) } if eval.Type != structs.JobTypeService { t.Fatalf("bad: %#v", eval) } if eval.TriggeredBy != structs.EvalTriggerJobDeregister { t.Fatalf("bad: %#v", eval) } if eval.JobID != jobID { t.Fatalf("bad: %#v", eval) } if eval.JobModifyIndex != resp2.JobModifyIndex { t.Fatalf("bad: %#v", eval) } if eval.Status != structs.EvalStatusPending { t.Fatalf("bad: %#v", eval) } if eval.CreateTime == 0 { t.Fatalf("eval CreateTime is unset: %#v", eval) } if eval.ModifyTime == 0 { t.Fatalf("eval ModifyTime is unset: %#v", eval) } } func TestJobEndpoint_Deregister_EvalPriority(t *testing.T) { t.Parallel() requireAssert := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.Canonicalize() // Register the job. requireAssert.NoError(msgpackrpc.CallWithCodec(codec, "Job.Register", &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, }, &structs.JobRegisterResponse{})) // Create the deregister request. deregReq := &structs.JobDeregisterRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, EvalPriority: 99, } var deregResp structs.JobDeregisterResponse requireAssert.NoError(msgpackrpc.CallWithCodec(codec, "Job.Deregister", deregReq, &deregResp)) // Grab the eval from the state, and check its priority is as expected. out, err := s1.fsm.State().EvalByID(nil, deregResp.EvalID) requireAssert.NoError(err) requireAssert.Equal(99, out.Priority) } func TestJobEndpoint_Deregister_Periodic(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.PeriodicJob() reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } // Deregister dereg := &structs.JobDeregisterRequest{ JobID: job.ID, Purge: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.JobModifyIndex == 0 { t.Fatalf("bad index: %d", resp2.Index) } // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out != nil { t.Fatalf("unexpected job") } if resp.EvalID != "" { t.Fatalf("Deregister created an eval for a periodic job") } } func TestJobEndpoint_Deregister_ParameterizedJob(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.BatchJob() job.ParameterizedJob = &structs.ParameterizedJobConfig{} reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } // Deregister dereg := &structs.JobDeregisterRequest{ JobID: job.ID, Purge: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.JobModifyIndex == 0 { t.Fatalf("bad index: %d", resp2.Index) } // Check for the node in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out != nil { t.Fatalf("unexpected job") } if resp.EvalID != "" { t.Fatalf("Deregister created an eval for a parameterized job") } } // TestJobEndpoint_Deregister_EvalCreation_Modern asserts that job deregister creates an eval // atomically with the registration func TestJobEndpoint_Deregister_EvalCreation_Modern(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request t.Run("job de-registration always create evals", func(t *testing.T) { job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) dereg := &structs.JobDeregisterRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2) require.NoError(t, err) require.NotEmpty(t, resp2.EvalID) state := s1.fsm.State() eval, err := state.EvalByID(nil, resp2.EvalID) require.Nil(t, err) require.NotNil(t, eval) require.EqualValues(t, resp2.EvalCreateIndex, eval.CreateIndex) require.Nil(t, evalUpdateFromRaft(t, s1, eval.ID)) }) // Registering a parameterized job shouldn't create an eval t.Run("periodic jobs shouldn't create an eval", func(t *testing.T) { job := mock.PeriodicJob() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) require.NotZero(t, resp.Index) dereg := &structs.JobDeregisterRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2) require.NoError(t, err) require.Empty(t, resp2.EvalID) }) } // TestJobEndpoint_Deregister_EvalCreation_Legacy asserts that job deregister // creates an eval atomically with the registration, but handle legacy clients // by adding a new eval update func TestJobEndpoint_Deregister_EvalCreation_Legacy(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.BootstrapExpect = 2 c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() s2, cleanupS2 := TestServer(t, func(c *Config) { c.BootstrapExpect = 2 c.NumSchedulers = 0 // Prevent automatic dequeue // simulate presense of a server that doesn't handle // new registration eval c.Build = "0.12.0" }) defer cleanupS2() TestJoin(t, s1, s2) testutil.WaitForLeader(t, s1.RPC) testutil.WaitForLeader(t, s2.RPC) // keep s1 as the leader if leader, _ := s1.getLeader(); !leader { s1, s2 = s2, s1 } codec := rpcClient(t, s1) // Create the register request t.Run("job registration always create evals", func(t *testing.T) { job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) dereg := &structs.JobDeregisterRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2) require.NoError(t, err) require.NotEmpty(t, resp2.EvalID) state := s1.fsm.State() eval, err := state.EvalByID(nil, resp2.EvalID) require.Nil(t, err) require.NotNil(t, eval) require.EqualValues(t, resp2.EvalCreateIndex, eval.CreateIndex) raftEval := evalUpdateFromRaft(t, s1, eval.ID) require.Equal(t, eval, raftEval) }) // Registering a parameterized job shouldn't create an eval t.Run("periodic jobs shouldn't create an eval", func(t *testing.T) { job := mock.PeriodicJob() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp) require.NoError(t, err) require.NotZero(t, resp.Index) dereg := &structs.JobDeregisterRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp2) require.NoError(t, err) require.Empty(t, resp2.EvalID) }) } func TestJobEndpoint_Deregister_NoShutdownDelay(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register requests job := mock.Job() reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp0 structs.JobRegisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp0)) // Deregister but don't purge dereg1 := &structs.JobDeregisterRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp1 structs.JobDeregisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg1, &resp1)) require.NotZero(resp1.Index) // Check for the job in the FSM state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.NoError(err) require.NotNil(out) require.True(out.Stop) // Lookup the evaluation eval, err := state.EvalByID(nil, resp1.EvalID) require.NoError(err) require.NotNil(eval) require.EqualValues(resp1.EvalCreateIndex, eval.CreateIndex) require.Equal(structs.EvalTriggerJobDeregister, eval.TriggeredBy) // Lookup allocation transitions var ws memdb.WatchSet allocs, err := state.AllocsByJob(ws, job.Namespace, job.ID, true) require.NoError(err) for _, alloc := range allocs { require.Nil(alloc.DesiredTransition) } // Deregister with no shutdown delay dereg2 := &structs.JobDeregisterRequest{ JobID: job.ID, NoShutdownDelay: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobDeregisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg2, &resp2)) require.NotZero(resp2.Index) // Lookup the evaluation eval, err = state.EvalByID(nil, resp2.EvalID) require.NoError(err) require.NotNil(eval) require.EqualValues(resp2.EvalCreateIndex, eval.CreateIndex) require.Equal(structs.EvalTriggerJobDeregister, eval.TriggeredBy) // Lookup allocation transitions allocs, err = state.AllocsByJob(ws, job.Namespace, job.ID, true) require.NoError(err) for _, alloc := range allocs { require.NotNil(alloc.DesiredTransition) require.True(*(alloc.DesiredTransition.NoShutdownDelay)) } } func TestJobEndpoint_BatchDeregister(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register requests job := mock.Job() reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp)) job2 := mock.Job() job2.Priority = 1 reg2 := &structs.JobRegisterRequest{ Job: job2, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job2.Namespace, }, } // Fetch the response require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Register", reg2, &resp)) // Deregister dereg := &structs.JobBatchDeregisterRequest{ Jobs: map[structs.NamespacedID]*structs.JobDeregisterOptions{ { ID: job.ID, Namespace: job.Namespace, }: {}, { ID: job2.ID, Namespace: job2.Namespace, }: { Purge: true, }, }, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobBatchDeregisterResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.BatchDeregister", dereg, &resp2)) require.NotZero(resp2.Index) // Check for the job in the FSM state := s1.fsm.State() out, err := state.JobByID(nil, job.Namespace, job.ID) require.Nil(err) require.NotNil(out) require.True(out.Stop) out, err = state.JobByID(nil, job2.Namespace, job2.ID) require.Nil(err) require.Nil(out) // Lookup the evaluation for jobNS, eval := range resp2.JobEvals { expectedJob := job if jobNS.ID != job.ID { expectedJob = job2 } eval, err := state.EvalByID(nil, eval) require.Nil(err) require.NotNil(eval) require.EqualValues(resp2.Index, eval.CreateIndex) require.Equal(expectedJob.Priority, eval.Priority) require.Equal(expectedJob.Type, eval.Type) require.Equal(structs.EvalTriggerJobDeregister, eval.TriggeredBy) require.Equal(expectedJob.ID, eval.JobID) require.Equal(structs.EvalStatusPending, eval.Status) require.NotZero(eval.CreateTime) require.NotZero(eval.ModifyTime) } } func TestJobEndpoint_BatchDeregister_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create and register a job job, job2 := mock.Job(), mock.Job() require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 100, job)) require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 101, job2)) // Deregister req := &structs.JobBatchDeregisterRequest{ Jobs: map[structs.NamespacedID]*structs.JobDeregisterOptions{ { ID: job.ID, Namespace: job.Namespace, }: {}, { ID: job2.ID, Namespace: job2.Namespace, }: {}, }, WriteRequest: structs.WriteRequest{ Region: "global", }, } // Expect failure for request without a token var resp structs.JobBatchDeregisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.BatchDeregister", req, &resp) require.NotNil(err) require.True(structs.IsErrPermissionDenied(err)) // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) req.AuthToken = invalidToken.SecretID var invalidResp structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.BatchDeregister", req, &invalidResp) require.NotNil(err) require.True(structs.IsErrPermissionDenied(err)) // Expect success with a valid management token req.AuthToken = root.SecretID var validResp structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.BatchDeregister", req, &validResp) require.Nil(err) require.NotEqual(validResp.Index, 0) // Expect success with a valid token validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilitySubmitJob})) req.AuthToken = validToken.SecretID var validResp2 structs.JobDeregisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.BatchDeregister", req, &validResp2) require.Nil(err) require.NotEqual(validResp2.Index, 0) } func TestJobEndpoint_Deregister_Priority(t *testing.T) { t.Parallel() requireAssertion := require.New(t) s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) fsmState := s1.fsm.State() // Create a job which a custom priority and register this. job := mock.Job() job.Priority = 90 err := fsmState.UpsertJob(structs.MsgTypeTestSetup, 100, job) requireAssertion.Nil(err) // Deregister. dereg := &structs.JobDeregisterRequest{ JobID: job.ID, Purge: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobDeregisterResponse requireAssertion.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deregister", dereg, &resp)) requireAssertion.NotZero(resp.Index) // Check for the job in the FSM which should not be there as it was purged. out, err := fsmState.JobByID(nil, job.Namespace, job.ID) requireAssertion.Nil(err) requireAssertion.Nil(out) // Lookup the evaluation eval, err := fsmState.EvalByID(nil, resp.EvalID) requireAssertion.Nil(err) requireAssertion.NotNil(eval) requireAssertion.EqualValues(resp.EvalCreateIndex, eval.CreateIndex) requireAssertion.Equal(job.Priority, eval.Priority) requireAssertion.Equal(job.Type, eval.Type) requireAssertion.Equal(structs.EvalTriggerJobDeregister, eval.TriggeredBy) requireAssertion.Equal(job.ID, eval.JobID) requireAssertion.Equal(structs.EvalStatusPending, eval.Status) requireAssertion.NotZero(eval.CreateTime) requireAssertion.NotZero(eval.ModifyTime) } func TestJobEndpoint_GetJob(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } job.CreateIndex = resp.JobModifyIndex job.ModifyIndex = resp.JobModifyIndex job.JobModifyIndex = resp.JobModifyIndex // Lookup the job get := &structs.JobSpecificRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.SingleJobResponse if err := msgpackrpc.CallWithCodec(codec, "Job.GetJob", get, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.Index != resp.JobModifyIndex { t.Fatalf("Bad index: %d %d", resp2.Index, resp.Index) } // Make a copy of the origin job and change the service name so that we can // do a deep equal with the response from the GET JOB Api j := job j.TaskGroups[0].Tasks[0].Services[0].Name = "web-frontend" for tgix, tg := range j.TaskGroups { for tidx, t := range tg.Tasks { for sidx, service := range t.Services { for cidx, check := range service.Checks { check.Name = resp2.Job.TaskGroups[tgix].Tasks[tidx].Services[sidx].Checks[cidx].Name } } } } // Clear the submit times j.SubmitTime = 0 resp2.Job.SubmitTime = 0 if !reflect.DeepEqual(j, resp2.Job) { t.Fatalf("bad: %#v %#v", job, resp2.Job) } // Lookup non-existing job get.JobID = "foobarbaz" if err := msgpackrpc.CallWithCodec(codec, "Job.GetJob", get, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.Index != resp.JobModifyIndex { t.Fatalf("Bad index: %d %d", resp2.Index, resp.Index) } if resp2.Job != nil { t.Fatalf("unexpected job") } } func TestJobEndpoint_GetJob_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create the job job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) // Lookup the job get := &structs.JobSpecificRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } // Looking up the job without a token should fail var resp structs.SingleJobResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJob", get, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) get.AuthToken = invalidToken.SecretID var invalidResp structs.SingleJobResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJob", get, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Looking up the job with a management token should succeed get.AuthToken = root.SecretID var validResp structs.SingleJobResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJob", get, &validResp) require.Nil(err) require.Equal(job.ID, validResp.Job.ID) // Looking up the job with a valid token should succeed validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) get.AuthToken = validToken.SecretID var validResp2 structs.SingleJobResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJob", get, &validResp2) require.Nil(err) require.Equal(job.ID, validResp2.Job.ID) } func TestJobEndpoint_GetJob_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() state := s1.fsm.State() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the jobs job1 := mock.Job() job2 := mock.Job() // Upsert a job we are not interested in first. time.AfterFunc(100*time.Millisecond, func() { if err := state.UpsertJob(structs.MsgTypeTestSetup, 100, job1); err != nil { t.Fatalf("err: %v", err) } }) // Upsert another job later which should trigger the watch. time.AfterFunc(200*time.Millisecond, func() { if err := state.UpsertJob(structs.MsgTypeTestSetup, 200, job2); err != nil { t.Fatalf("err: %v", err) } }) req := &structs.JobSpecificRequest{ JobID: job2.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job2.Namespace, MinQueryIndex: 150, }, } start := time.Now() var resp structs.SingleJobResponse if err := msgpackrpc.CallWithCodec(codec, "Job.GetJob", req, &resp); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } if resp.Index != 200 { t.Fatalf("Bad index: %d %d", resp.Index, 200) } if resp.Job == nil || resp.Job.ID != job2.ID { t.Fatalf("bad: %#v", resp.Job) } // Job delete fires watches time.AfterFunc(100*time.Millisecond, func() { if err := state.DeleteJob(300, job2.Namespace, job2.ID); err != nil { t.Fatalf("err: %v", err) } }) req.QueryOptions.MinQueryIndex = 250 start = time.Now() var resp2 structs.SingleJobResponse if err := msgpackrpc.CallWithCodec(codec, "Job.GetJob", req, &resp2); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 100*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp2) } if resp2.Index != 300 { t.Fatalf("Bad index: %d %d", resp2.Index, 300) } if resp2.Job != nil { t.Fatalf("bad: %#v", resp2.Job) } } func TestJobEndpoint_GetJobVersions(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.Priority = 88 reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } // Register the job again to create another version job.Priority = 100 if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } // Lookup the job get := &structs.JobVersionsRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } var versionsResp structs.JobVersionsResponse if err := msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", get, &versionsResp); err != nil { t.Fatalf("err: %v", err) } if versionsResp.Index != resp.JobModifyIndex { t.Fatalf("Bad index: %d %d", versionsResp.Index, resp.Index) } // Make sure there are two job versions versions := versionsResp.Versions if l := len(versions); l != 2 { t.Fatalf("Got %d versions; want 2", l) } if v := versions[0]; v.Priority != 100 || v.ID != job.ID || v.Version != 1 { t.Fatalf("bad: %+v", v) } if v := versions[1]; v.Priority != 88 || v.ID != job.ID || v.Version != 0 { t.Fatalf("bad: %+v", v) } // Lookup non-existing job get.JobID = "foobarbaz" if err := msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", get, &versionsResp); err != nil { t.Fatalf("err: %v", err) } if versionsResp.Index != resp.JobModifyIndex { t.Fatalf("Bad index: %d %d", versionsResp.Index, resp.Index) } if l := len(versionsResp.Versions); l != 0 { t.Fatalf("unexpected versions: %d", l) } } func TestJobEndpoint_GetJobVersions_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create two versions of a job with different priorities job := mock.Job() job.Priority = 88 err := state.UpsertJob(structs.MsgTypeTestSetup, 10, job) require.Nil(err) job.Priority = 100 err = state.UpsertJob(structs.MsgTypeTestSetup, 100, job) require.Nil(err) // Lookup the job get := &structs.JobVersionsRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } // Attempt to fetch without a token should fail var resp structs.JobVersionsResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", get, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) get.AuthToken = invalidToken.SecretID var invalidResp structs.JobVersionsResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", get, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Expect success for request with a valid management token get.AuthToken = root.SecretID var validResp structs.JobVersionsResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", get, &validResp) require.Nil(err) // Expect success for request with a valid token validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) get.AuthToken = validToken.SecretID var validResp2 structs.JobVersionsResponse err = msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", get, &validResp2) require.Nil(err) // Make sure there are two job versions versions := validResp2.Versions require.Equal(2, len(versions)) require.Equal(versions[0].ID, job.ID) require.Equal(versions[1].ID, job.ID) } func TestJobEndpoint_GetJobVersions_Diff(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() job.Priority = 88 reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } // Register the job again to create another version job.Priority = 90 if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } // Register the job again to create another version job.Priority = 100 if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } // Lookup the job get := &structs.JobVersionsRequest{ JobID: job.ID, Diffs: true, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } var versionsResp structs.JobVersionsResponse if err := msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", get, &versionsResp); err != nil { t.Fatalf("err: %v", err) } if versionsResp.Index != resp.JobModifyIndex { t.Fatalf("Bad index: %d %d", versionsResp.Index, resp.Index) } // Make sure there are two job versions versions := versionsResp.Versions if l := len(versions); l != 3 { t.Fatalf("Got %d versions; want 3", l) } if v := versions[0]; v.Priority != 100 || v.ID != job.ID || v.Version != 2 { t.Fatalf("bad: %+v", v) } if v := versions[1]; v.Priority != 90 || v.ID != job.ID || v.Version != 1 { t.Fatalf("bad: %+v", v) } if v := versions[2]; v.Priority != 88 || v.ID != job.ID || v.Version != 0 { t.Fatalf("bad: %+v", v) } // Ensure we got diffs diffs := versionsResp.Diffs if l := len(diffs); l != 2 { t.Fatalf("Got %d diffs; want 2", l) } d1 := diffs[0] if len(d1.Fields) != 1 { t.Fatalf("Got too many diffs: %#v", d1) } if d1.Fields[0].Name != "Priority" { t.Fatalf("Got wrong field: %#v", d1) } if d1.Fields[0].Old != "90" && d1.Fields[0].New != "100" { t.Fatalf("Got wrong field values: %#v", d1) } d2 := diffs[1] if len(d2.Fields) != 1 { t.Fatalf("Got too many diffs: %#v", d2) } if d2.Fields[0].Name != "Priority" { t.Fatalf("Got wrong field: %#v", d2) } if d2.Fields[0].Old != "88" && d1.Fields[0].New != "90" { t.Fatalf("Got wrong field values: %#v", d2) } } func TestJobEndpoint_GetJobVersions_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() state := s1.fsm.State() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the jobs job1 := mock.Job() job2 := mock.Job() job3 := mock.Job() job3.ID = job2.ID job3.Priority = 1 // Upsert a job we are not interested in first. time.AfterFunc(100*time.Millisecond, func() { if err := state.UpsertJob(structs.MsgTypeTestSetup, 100, job1); err != nil { t.Fatalf("err: %v", err) } }) // Upsert another job later which should trigger the watch. time.AfterFunc(200*time.Millisecond, func() { if err := state.UpsertJob(structs.MsgTypeTestSetup, 200, job2); err != nil { t.Fatalf("err: %v", err) } }) req := &structs.JobVersionsRequest{ JobID: job2.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job2.Namespace, MinQueryIndex: 150, }, } start := time.Now() var resp structs.JobVersionsResponse if err := msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", req, &resp); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } if resp.Index != 200 { t.Fatalf("Bad index: %d %d", resp.Index, 200) } if len(resp.Versions) != 1 || resp.Versions[0].ID != job2.ID { t.Fatalf("bad: %#v", resp.Versions) } // Upsert the job again which should trigger the watch. time.AfterFunc(100*time.Millisecond, func() { if err := state.UpsertJob(structs.MsgTypeTestSetup, 300, job3); err != nil { t.Fatalf("err: %v", err) } }) req2 := &structs.JobVersionsRequest{ JobID: job3.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job3.Namespace, MinQueryIndex: 250, }, } var resp2 structs.JobVersionsResponse start = time.Now() if err := msgpackrpc.CallWithCodec(codec, "Job.GetJobVersions", req2, &resp2); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 100*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } if resp2.Index != 300 { t.Fatalf("Bad index: %d %d", resp.Index, 300) } if len(resp2.Versions) != 2 { t.Fatalf("bad: %#v", resp2.Versions) } } func TestJobEndpoint_GetJobSummary(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", reg, &resp); err != nil { t.Fatalf("err: %v", err) } job.CreateIndex = resp.JobModifyIndex job.ModifyIndex = resp.JobModifyIndex job.JobModifyIndex = resp.JobModifyIndex // Lookup the job summary get := &structs.JobSummaryRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobSummaryResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Summary", get, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.Index != resp.JobModifyIndex { t.Fatalf("Bad index: %d %d", resp2.Index, resp.Index) } expectedJobSummary := structs.JobSummary{ JobID: job.ID, Namespace: job.Namespace, Summary: map[string]structs.TaskGroupSummary{ "web": {}, }, Children: new(structs.JobChildrenSummary), CreateIndex: job.CreateIndex, ModifyIndex: job.CreateIndex, } if !reflect.DeepEqual(resp2.JobSummary, &expectedJobSummary) { t.Fatalf("expected: %v, actual: %v", expectedJobSummary, resp2.JobSummary) } } func TestJobEndpoint_Summary_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the job job := mock.Job() reg := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } reg.AuthToken = root.SecretID var err error // Register the job with a valid token var regResp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Register", reg, ®Resp) require.Nil(err) job.CreateIndex = regResp.JobModifyIndex job.ModifyIndex = regResp.JobModifyIndex job.JobModifyIndex = regResp.JobModifyIndex req := &structs.JobSummaryRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } // Expect failure for request without a token var resp structs.JobSummaryResponse err = msgpackrpc.CallWithCodec(codec, "Job.Summary", req, &resp) require.NotNil(err) expectedJobSummary := &structs.JobSummary{ JobID: job.ID, Namespace: job.Namespace, Summary: map[string]structs.TaskGroupSummary{ "web": {}, }, Children: new(structs.JobChildrenSummary), CreateIndex: job.CreateIndex, ModifyIndex: job.ModifyIndex, } // Expect success when using a management token req.AuthToken = root.SecretID var mgmtResp structs.JobSummaryResponse err = msgpackrpc.CallWithCodec(codec, "Job.Summary", req, &mgmtResp) require.Nil(err) require.Equal(expectedJobSummary, mgmtResp.JobSummary) // Create the namespace policy and tokens state := s1.fsm.State() // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) req.AuthToken = invalidToken.SecretID var invalidResp structs.JobSummaryResponse err = msgpackrpc.CallWithCodec(codec, "Job.Summary", req, &invalidResp) require.NotNil(err) // Try with a valid token validToken := mock.CreatePolicyAndToken(t, state, 1001, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) req.AuthToken = validToken.SecretID var authResp structs.JobSummaryResponse err = msgpackrpc.CallWithCodec(codec, "Job.Summary", req, &authResp) require.Nil(err) require.Equal(expectedJobSummary, authResp.JobSummary) } func TestJobEndpoint_GetJobSummary_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() state := s1.fsm.State() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create a job and insert it job1 := mock.Job() time.AfterFunc(200*time.Millisecond, func() { if err := state.UpsertJob(structs.MsgTypeTestSetup, 100, job1); err != nil { t.Fatalf("err: %v", err) } }) // Ensure the job summary request gets fired req := &structs.JobSummaryRequest{ JobID: job1.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job1.Namespace, MinQueryIndex: 50, }, } var resp structs.JobSummaryResponse start := time.Now() if err := msgpackrpc.CallWithCodec(codec, "Job.Summary", req, &resp); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } // Upsert an allocation for the job which should trigger the watch. time.AfterFunc(200*time.Millisecond, func() { alloc := mock.Alloc() alloc.JobID = job1.ID alloc.Job = job1 if err := state.UpsertAllocs(structs.MsgTypeTestSetup, 200, []*structs.Allocation{alloc}); err != nil { t.Fatalf("err: %v", err) } }) req = &structs.JobSummaryRequest{ JobID: job1.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job1.Namespace, MinQueryIndex: 199, }, } start = time.Now() var resp1 structs.JobSummaryResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Summary", req, &resp1); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } if resp1.Index != 200 { t.Fatalf("Bad index: %d %d", resp.Index, 200) } if resp1.JobSummary == nil { t.Fatalf("bad: %#v", resp) } // Job delete fires watches time.AfterFunc(100*time.Millisecond, func() { if err := state.DeleteJob(300, job1.Namespace, job1.ID); err != nil { t.Fatalf("err: %v", err) } }) req.QueryOptions.MinQueryIndex = 250 start = time.Now() var resp2 structs.SingleJobResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Summary", req, &resp2); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 100*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp2) } if resp2.Index != 300 { t.Fatalf("Bad index: %d %d", resp2.Index, 300) } if resp2.Job != nil { t.Fatalf("bad: %#v", resp2.Job) } } func TestJobEndpoint_ListJobs(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() state := s1.fsm.State() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.NoError(t, err) // Lookup the jobs get := &structs.JobListRequest{ QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } var resp2 structs.JobListResponse err = msgpackrpc.CallWithCodec(codec, "Job.List", get, &resp2) require.NoError(t, err) require.Equal(t, uint64(1000), resp2.Index) require.Len(t, resp2.Jobs, 1) require.Equal(t, job.ID, resp2.Jobs[0].ID) require.Equal(t, job.Namespace, resp2.Jobs[0].Namespace) // Lookup the jobs by prefix get = &structs.JobListRequest{ QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, Prefix: resp2.Jobs[0].ID[:4], }, } var resp3 structs.JobListResponse err = msgpackrpc.CallWithCodec(codec, "Job.List", get, &resp3) require.NoError(t, err) require.Equal(t, uint64(1000), resp3.Index) require.Len(t, resp3.Jobs, 1) require.Equal(t, job.ID, resp3.Jobs[0].ID) require.Equal(t, job.Namespace, resp3.Jobs[0].Namespace) } // TestJobEndpoint_ListJobs_AllNamespaces_OSS asserts that server // returns all jobs across namespace. // func TestJobEndpoint_ListJobs_AllNamespaces_OSS(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() state := s1.fsm.State() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) if err != nil { t.Fatalf("err: %v", err) } // Lookup the jobs get := &structs.JobListRequest{ QueryOptions: structs.QueryOptions{ Region: "global", Namespace: "*", }, } var resp2 structs.JobListResponse err = msgpackrpc.CallWithCodec(codec, "Job.List", get, &resp2) require.NoError(t, err) require.Equal(t, uint64(1000), resp2.Index) require.Len(t, resp2.Jobs, 1) require.Equal(t, job.ID, resp2.Jobs[0].ID) require.Equal(t, structs.DefaultNamespace, resp2.Jobs[0].Namespace) // Lookup the jobs by prefix get = &structs.JobListRequest{ QueryOptions: structs.QueryOptions{ Region: "global", Namespace: "*", Prefix: resp2.Jobs[0].ID[:4], }, } var resp3 structs.JobListResponse err = msgpackrpc.CallWithCodec(codec, "Job.List", get, &resp3) require.NoError(t, err) require.Equal(t, uint64(1000), resp3.Index) require.Len(t, resp3.Jobs, 1) require.Equal(t, job.ID, resp3.Jobs[0].ID) require.Equal(t, structs.DefaultNamespace, resp2.Jobs[0].Namespace) // Lookup the jobs by prefix get = &structs.JobListRequest{ QueryOptions: structs.QueryOptions{ Region: "global", Namespace: "*", Prefix: "z" + resp2.Jobs[0].ID[:4], }, } var resp4 structs.JobListResponse err = msgpackrpc.CallWithCodec(codec, "Job.List", get, &resp4) require.NoError(t, err) require.Equal(t, uint64(1000), resp4.Index) require.Empty(t, resp4.Jobs) } func TestJobEndpoint_ListJobs_WithACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() var err error // Create the register request job := mock.Job() err = state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) req := &structs.JobListRequest{ QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } // Expect failure for request without a token var resp structs.JobListResponse err = msgpackrpc.CallWithCodec(codec, "Job.List", req, &resp) require.NotNil(err) // Expect success for request with a management token var mgmtResp structs.JobListResponse req.AuthToken = root.SecretID err = msgpackrpc.CallWithCodec(codec, "Job.List", req, &mgmtResp) require.Nil(err) require.Equal(1, len(mgmtResp.Jobs)) require.Equal(job.ID, mgmtResp.Jobs[0].ID) // Expect failure for request with a token that has incorrect permissions invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) req.AuthToken = invalidToken.SecretID var invalidResp structs.JobListResponse err = msgpackrpc.CallWithCodec(codec, "Job.List", req, &invalidResp) require.NotNil(err) // Try with a valid token with correct permissions validToken := mock.CreatePolicyAndToken(t, state, 1001, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) var validResp structs.JobListResponse req.AuthToken = validToken.SecretID err = msgpackrpc.CallWithCodec(codec, "Job.List", req, &validResp) require.Nil(err) require.Equal(1, len(validResp.Jobs)) require.Equal(job.ID, validResp.Jobs[0].ID) } func TestJobEndpoint_ListJobs_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() state := s1.fsm.State() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the job job := mock.Job() // Upsert job triggers watches time.AfterFunc(100*time.Millisecond, func() { if err := state.UpsertJob(structs.MsgTypeTestSetup, 100, job); err != nil { t.Fatalf("err: %v", err) } }) req := &structs.JobListRequest{ QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, MinQueryIndex: 50, }, } start := time.Now() var resp structs.JobListResponse if err := msgpackrpc.CallWithCodec(codec, "Job.List", req, &resp); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 100*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } if resp.Index != 100 { t.Fatalf("Bad index: %d %d", resp.Index, 100) } if len(resp.Jobs) != 1 || resp.Jobs[0].ID != job.ID { t.Fatalf("bad: %#v", resp) } // Job deletion triggers watches time.AfterFunc(100*time.Millisecond, func() { if err := state.DeleteJob(200, job.Namespace, job.ID); err != nil { t.Fatalf("err: %v", err) } }) req.MinQueryIndex = 150 start = time.Now() var resp2 structs.JobListResponse if err := msgpackrpc.CallWithCodec(codec, "Job.List", req, &resp2); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 100*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp2) } if resp2.Index != 200 { t.Fatalf("Bad index: %d %d", resp2.Index, 200) } if len(resp2.Jobs) != 0 { t.Fatalf("bad: %#v", resp2) } } func TestJobEndpoint_Allocations(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request alloc1 := mock.Alloc() alloc2 := mock.Alloc() alloc2.JobID = alloc1.JobID state := s1.fsm.State() state.UpsertJobSummary(998, mock.JobSummary(alloc1.JobID)) state.UpsertJobSummary(999, mock.JobSummary(alloc2.JobID)) err := state.UpsertAllocs(structs.MsgTypeTestSetup, 1000, []*structs.Allocation{alloc1, alloc2}) if err != nil { t.Fatalf("err: %v", err) } // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: alloc1.JobID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: alloc1.Job.Namespace, }, } var resp2 structs.JobAllocationsResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Allocations", get, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.Index != 1000 { t.Fatalf("Bad index: %d %d", resp2.Index, 1000) } if len(resp2.Allocations) != 2 { t.Fatalf("bad: %#v", resp2.Allocations) } } func TestJobEndpoint_Allocations_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create allocations for a job alloc1 := mock.Alloc() alloc2 := mock.Alloc() alloc2.JobID = alloc1.JobID state.UpsertJobSummary(998, mock.JobSummary(alloc1.JobID)) state.UpsertJobSummary(999, mock.JobSummary(alloc2.JobID)) err := state.UpsertAllocs(structs.MsgTypeTestSetup, 1000, []*structs.Allocation{alloc1, alloc2}) require.Nil(err) // Look up allocations for that job get := &structs.JobSpecificRequest{ JobID: alloc1.JobID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: alloc1.Job.Namespace, }, } // Attempt to fetch the response without a token should fail var resp structs.JobAllocationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Allocations", get, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with an invalid token should fail invalidToken := mock.CreatePolicyAndToken(t, state, 1001, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) get.AuthToken = invalidToken.SecretID var invalidResp structs.JobAllocationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Allocations", get, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with valid management token should succeed get.AuthToken = root.SecretID var validResp structs.JobAllocationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Allocations", get, &validResp) require.Nil(err) // Attempt to fetch the response with valid management token should succeed validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) get.AuthToken = validToken.SecretID var validResp2 structs.JobAllocationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Allocations", get, &validResp2) require.Nil(err) require.Equal(2, len(validResp2.Allocations)) } func TestJobEndpoint_Allocations_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request alloc1 := mock.Alloc() alloc2 := mock.Alloc() alloc2.JobID = "job1" state := s1.fsm.State() // First upsert an unrelated alloc time.AfterFunc(100*time.Millisecond, func() { state.UpsertJobSummary(99, mock.JobSummary(alloc1.JobID)) err := state.UpsertAllocs(structs.MsgTypeTestSetup, 100, []*structs.Allocation{alloc1}) if err != nil { t.Fatalf("err: %v", err) } }) // Upsert an alloc for the job we are interested in later time.AfterFunc(200*time.Millisecond, func() { state.UpsertJobSummary(199, mock.JobSummary(alloc2.JobID)) err := state.UpsertAllocs(structs.MsgTypeTestSetup, 200, []*structs.Allocation{alloc2}) if err != nil { t.Fatalf("err: %v", err) } }) // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: "job1", QueryOptions: structs.QueryOptions{ Region: "global", Namespace: alloc1.Job.Namespace, MinQueryIndex: 150, }, } var resp structs.JobAllocationsResponse start := time.Now() if err := msgpackrpc.CallWithCodec(codec, "Job.Allocations", get, &resp); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } if resp.Index != 200 { t.Fatalf("Bad index: %d %d", resp.Index, 200) } if len(resp.Allocations) != 1 || resp.Allocations[0].JobID != "job1" { t.Fatalf("bad: %#v", resp.Allocations) } } // TestJobEndpoint_Allocations_NoJobID asserts not setting a JobID in the // request returns an error. func TestJobEndpoint_Allocations_NoJobID(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) get := &structs.JobSpecificRequest{ JobID: "", QueryOptions: structs.QueryOptions{ Region: "global", Namespace: structs.DefaultNamespace, }, } var resp structs.JobAllocationsResponse err := msgpackrpc.CallWithCodec(codec, "Job.Allocations", get, &resp) require.Error(t, err) require.Contains(t, err.Error(), "missing job ID") } func TestJobEndpoint_Evaluations(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request eval1 := mock.Eval() eval2 := mock.Eval() eval2.JobID = eval1.JobID state := s1.fsm.State() err := state.UpsertEvals(structs.MsgTypeTestSetup, 1000, []*structs.Evaluation{eval1, eval2}) if err != nil { t.Fatalf("err: %v", err) } // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: eval1.JobID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: eval1.Namespace, }, } var resp2 structs.JobEvaluationsResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Evaluations", get, &resp2); err != nil { t.Fatalf("err: %v", err) } if resp2.Index != 1000 { t.Fatalf("Bad index: %d %d", resp2.Index, 1000) } if len(resp2.Evaluations) != 2 { t.Fatalf("bad: %#v", resp2.Evaluations) } } func TestJobEndpoint_Evaluations_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create evaluations for the same job eval1 := mock.Eval() eval2 := mock.Eval() eval2.JobID = eval1.JobID err := state.UpsertEvals(structs.MsgTypeTestSetup, 1000, []*structs.Evaluation{eval1, eval2}) require.Nil(err) // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: eval1.JobID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: eval1.Namespace, }, } // Attempt to fetch without providing a token var resp structs.JobEvaluationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluations", get, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1001, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) get.AuthToken = invalidToken.SecretID var invalidResp structs.JobEvaluationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluations", get, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch with valid management token should succeed get.AuthToken = root.SecretID var validResp structs.JobEvaluationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluations", get, &validResp) require.Nil(err) require.Equal(2, len(validResp.Evaluations)) // Attempt to fetch with valid token should succeed validToken := mock.CreatePolicyAndToken(t, state, 1003, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) get.AuthToken = validToken.SecretID var validResp2 structs.JobEvaluationsResponse err = msgpackrpc.CallWithCodec(codec, "Job.Evaluations", get, &validResp2) require.Nil(err) require.Equal(2, len(validResp2.Evaluations)) } func TestJobEndpoint_Evaluations_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request eval1 := mock.Eval() eval2 := mock.Eval() eval2.JobID = "job1" state := s1.fsm.State() // First upsert an unrelated eval time.AfterFunc(100*time.Millisecond, func() { err := state.UpsertEvals(structs.MsgTypeTestSetup, 100, []*structs.Evaluation{eval1}) if err != nil { t.Fatalf("err: %v", err) } }) // Upsert an eval for the job we are interested in later time.AfterFunc(200*time.Millisecond, func() { err := state.UpsertEvals(structs.MsgTypeTestSetup, 200, []*structs.Evaluation{eval2}) if err != nil { t.Fatalf("err: %v", err) } }) // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: "job1", QueryOptions: structs.QueryOptions{ Region: "global", Namespace: eval1.Namespace, MinQueryIndex: 150, }, } var resp structs.JobEvaluationsResponse start := time.Now() if err := msgpackrpc.CallWithCodec(codec, "Job.Evaluations", get, &resp); err != nil { t.Fatalf("err: %v", err) } if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } if resp.Index != 200 { t.Fatalf("Bad index: %d %d", resp.Index, 200) } if len(resp.Evaluations) != 1 || resp.Evaluations[0].JobID != "job1" { t.Fatalf("bad: %#v", resp.Evaluations) } } func TestJobEndpoint_Deployments(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() require := require.New(t) // Create the register request j := mock.Job() d1 := mock.Deployment() d2 := mock.Deployment() d1.JobID = j.ID d2.JobID = j.ID require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 1000, j), "UpsertJob") d1.JobCreateIndex = j.CreateIndex d2.JobCreateIndex = j.CreateIndex require.Nil(state.UpsertDeployment(1001, d1), "UpsertDeployment") require.Nil(state.UpsertDeployment(1002, d2), "UpsertDeployment") // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: j.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: j.Namespace, }, } var resp structs.DeploymentListResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deployments", get, &resp), "RPC") require.EqualValues(1002, resp.Index, "response index") require.Len(resp.Deployments, 2, "deployments for job") } func TestJobEndpoint_Deployments_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create a job and corresponding deployments j := mock.Job() d1 := mock.Deployment() d2 := mock.Deployment() d1.JobID = j.ID d2.JobID = j.ID require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 1000, j), "UpsertJob") d1.JobCreateIndex = j.CreateIndex d2.JobCreateIndex = j.CreateIndex require.Nil(state.UpsertDeployment(1001, d1), "UpsertDeployment") require.Nil(state.UpsertDeployment(1002, d2), "UpsertDeployment") // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: j.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: j.Namespace, }, } // Lookup with no token should fail var resp structs.DeploymentListResponse err := msgpackrpc.CallWithCodec(codec, "Job.Deployments", get, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1001, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) get.AuthToken = invalidToken.SecretID var invalidResp structs.DeploymentListResponse err = msgpackrpc.CallWithCodec(codec, "Job.Deployments", get, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Lookup with valid management token should succeed get.AuthToken = root.SecretID var validResp structs.DeploymentListResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deployments", get, &validResp), "RPC") require.EqualValues(1002, validResp.Index, "response index") require.Len(validResp.Deployments, 2, "deployments for job") // Lookup with valid token should succeed validToken := mock.CreatePolicyAndToken(t, state, 1005, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) get.AuthToken = validToken.SecretID var validResp2 structs.DeploymentListResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deployments", get, &validResp2), "RPC") require.EqualValues(1002, validResp2.Index, "response index") require.Len(validResp2.Deployments, 2, "deployments for job") } func TestJobEndpoint_Deployments_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() require := require.New(t) // Create the register request j := mock.Job() d1 := mock.Deployment() d2 := mock.Deployment() d2.JobID = j.ID require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 50, j), "UpsertJob") d2.JobCreateIndex = j.CreateIndex // First upsert an unrelated eval time.AfterFunc(100*time.Millisecond, func() { require.Nil(state.UpsertDeployment(100, d1), "UpsertDeployment") }) // Upsert an eval for the job we are interested in later time.AfterFunc(200*time.Millisecond, func() { require.Nil(state.UpsertDeployment(200, d2), "UpsertDeployment") }) // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: d2.JobID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: d2.Namespace, MinQueryIndex: 150, }, } var resp structs.DeploymentListResponse start := time.Now() require.Nil(msgpackrpc.CallWithCodec(codec, "Job.Deployments", get, &resp), "RPC") require.EqualValues(200, resp.Index, "response index") require.Len(resp.Deployments, 1, "deployments for job") require.Equal(d2.ID, resp.Deployments[0].ID, "returned deployment") if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } } func TestJobEndpoint_LatestDeployment(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() require := require.New(t) // Create the register request j := mock.Job() d1 := mock.Deployment() d2 := mock.Deployment() d1.JobID = j.ID d2.JobID = j.ID d2.CreateIndex = d1.CreateIndex + 100 d2.ModifyIndex = d2.CreateIndex + 100 require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 1000, j), "UpsertJob") d1.JobCreateIndex = j.CreateIndex d2.JobCreateIndex = j.CreateIndex require.Nil(state.UpsertDeployment(1001, d1), "UpsertDeployment") require.Nil(state.UpsertDeployment(1002, d2), "UpsertDeployment") // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: j.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: j.Namespace, }, } var resp structs.SingleDeploymentResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.LatestDeployment", get, &resp), "RPC") require.EqualValues(1002, resp.Index, "response index") require.NotNil(resp.Deployment, "want a deployment") require.Equal(d2.ID, resp.Deployment.ID, "latest deployment for job") } func TestJobEndpoint_LatestDeployment_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create a job and deployments j := mock.Job() d1 := mock.Deployment() d2 := mock.Deployment() d1.JobID = j.ID d2.JobID = j.ID d2.CreateIndex = d1.CreateIndex + 100 d2.ModifyIndex = d2.CreateIndex + 100 require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 1000, j), "UpsertJob") d1.JobCreateIndex = j.CreateIndex d2.JobCreateIndex = j.CreateIndex require.Nil(state.UpsertDeployment(1001, d1), "UpsertDeployment") require.Nil(state.UpsertDeployment(1002, d2), "UpsertDeployment") // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: j.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: j.Namespace, }, } // Attempt to fetch the response without a token should fail var resp structs.SingleDeploymentResponse err := msgpackrpc.CallWithCodec(codec, "Job.LatestDeployment", get, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with an invalid token should fail invalidToken := mock.CreatePolicyAndToken(t, state, 1001, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) get.AuthToken = invalidToken.SecretID var invalidResp structs.SingleDeploymentResponse err = msgpackrpc.CallWithCodec(codec, "Job.LatestDeployment", get, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Fetching latest deployment with a valid management token should succeed get.AuthToken = root.SecretID var validResp structs.SingleDeploymentResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.LatestDeployment", get, &validResp), "RPC") require.EqualValues(1002, validResp.Index, "response index") require.NotNil(validResp.Deployment, "want a deployment") require.Equal(d2.ID, validResp.Deployment.ID, "latest deployment for job") // Fetching latest deployment with a valid token should succeed validToken := mock.CreatePolicyAndToken(t, state, 1004, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})) get.AuthToken = validToken.SecretID var validResp2 structs.SingleDeploymentResponse require.Nil(msgpackrpc.CallWithCodec(codec, "Job.LatestDeployment", get, &validResp2), "RPC") require.EqualValues(1002, validResp2.Index, "response index") require.NotNil(validResp2.Deployment, "want a deployment") require.Equal(d2.ID, validResp2.Deployment.ID, "latest deployment for job") } func TestJobEndpoint_LatestDeployment_Blocking(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() require := require.New(t) // Create the register request j := mock.Job() d1 := mock.Deployment() d2 := mock.Deployment() d2.JobID = j.ID require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 50, j), "UpsertJob") d2.JobCreateIndex = j.CreateIndex // First upsert an unrelated eval time.AfterFunc(100*time.Millisecond, func() { require.Nil(state.UpsertDeployment(100, d1), "UpsertDeployment") }) // Upsert an eval for the job we are interested in later time.AfterFunc(200*time.Millisecond, func() { require.Nil(state.UpsertDeployment(200, d2), "UpsertDeployment") }) // Lookup the jobs get := &structs.JobSpecificRequest{ JobID: d2.JobID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: d2.Namespace, MinQueryIndex: 150, }, } var resp structs.SingleDeploymentResponse start := time.Now() require.Nil(msgpackrpc.CallWithCodec(codec, "Job.LatestDeployment", get, &resp), "RPC") require.EqualValues(200, resp.Index, "response index") require.NotNil(resp.Deployment, "deployment for job") require.Equal(d2.ID, resp.Deployment.ID, "returned deployment") if elapsed := time.Since(start); elapsed < 200*time.Millisecond { t.Fatalf("should block (returned in %s) %#v", elapsed, resp) } } func TestJobEndpoint_Plan_ACL(t *testing.T) { t.Parallel() s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create a plan request job := mock.Job() planReq := &structs.JobPlanRequest{ Job: job, Diff: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Try without a token, expect failure var planResp structs.JobPlanResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Plan", planReq, &planResp); err == nil { t.Fatalf("expected error") } // Try with a token planReq.AuthToken = root.SecretID if err := msgpackrpc.CallWithCodec(codec, "Job.Plan", planReq, &planResp); err != nil { t.Fatalf("err: %v", err) } } func TestJobEndpoint_Plan_WithDiff(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Create a plan request planReq := &structs.JobPlanRequest{ Job: job, Diff: true, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var planResp structs.JobPlanResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Plan", planReq, &planResp); err != nil { t.Fatalf("err: %v", err) } // Check the response if planResp.JobModifyIndex == 0 { t.Fatalf("bad cas: %d", planResp.JobModifyIndex) } if planResp.Annotations == nil { t.Fatalf("no annotations") } if planResp.Diff == nil { t.Fatalf("no diff") } if len(planResp.FailedTGAllocs) == 0 { t.Fatalf("no failed task group alloc metrics") } } func TestJobEndpoint_Plan_NoDiff(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request job := mock.Job() req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("err: %v", err) } if resp.Index == 0 { t.Fatalf("bad index: %d", resp.Index) } // Create a plan request planReq := &structs.JobPlanRequest{ Job: job, Diff: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var planResp structs.JobPlanResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Plan", planReq, &planResp); err != nil { t.Fatalf("err: %v", err) } // Check the response if planResp.JobModifyIndex == 0 { t.Fatalf("bad cas: %d", planResp.JobModifyIndex) } if planResp.Annotations == nil { t.Fatalf("no annotations") } if planResp.Diff != nil { t.Fatalf("got diff") } if len(planResp.FailedTGAllocs) == 0 { t.Fatalf("no failed task group alloc metrics") } } // TestJobEndpoint_Plan_Scaling asserts that the plan endpoint handles // jobs with scaling stanza func TestJobEndpoint_Plan_Scaling(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create a plan request job := mock.Job() tg := job.TaskGroups[0] tg.Tasks[0].Resources.MemoryMB = 999999999 scaling := &structs.ScalingPolicy{Min: 1, Max: 100, Type: structs.ScalingPolicyTypeHorizontal} tg.Scaling = scaling.TargetTaskGroup(job, tg) planReq := &structs.JobPlanRequest{ Job: job, Diff: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Try without a token, expect failure var planResp structs.JobPlanResponse err := msgpackrpc.CallWithCodec(codec, "Job.Plan", planReq, &planResp) require.NoError(t, err) require.NotEmpty(t, planResp.FailedTGAllocs) require.Contains(t, planResp.FailedTGAllocs, tg.Name) } func TestJobEndpoint_ImplicitConstraints_Vault(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Enable vault tr, f := true, false s1.config.VaultConfig.Enabled = &tr s1.config.VaultConfig.AllowUnauthenticated = &f // Replace the Vault Client on the server tvc := &TestVaultClient{} s1.vault = tvc policy := "foo" goodToken := uuid.Generate() goodPolicies := []string{"foo", "bar", "baz"} tvc.SetLookupTokenAllowedPolicies(goodToken, goodPolicies) // Create the register request with a job asking for a vault policy job := mock.Job() job.VaultToken = goodToken job.TaskGroups[0].Tasks[0].Vault = &structs.Vault{ Policies: []string{policy}, ChangeMode: structs.VaultChangeModeRestart, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("bad: %v", err) } // Check for the job in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } // Check that there is an implicit vault constraint constraints := out.TaskGroups[0].Constraints if len(constraints) != 1 { t.Fatalf("Expected an implicit constraint") } if !constraints[0].Equal(vaultConstraint) { t.Fatalf("Expected implicit vault constraint") } } func TestJobEndpoint_ValidateJob_ConsulConnect(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) validateJob := func(j *structs.Job) error { req := &structs.JobRegisterRequest{ Job: j, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: j.Namespace, }, } var resp structs.JobValidateResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Validate", req, &resp); err != nil { return err } if resp.Error != "" { return errors.New(resp.Error) } if len(resp.ValidationErrors) != 0 { return errors.New(strings.Join(resp.ValidationErrors, ",")) } if resp.Warnings != "" { return errors.New(resp.Warnings) } return nil } tgServices := []*structs.Service{ { Name: "count-api", PortLabel: "9001", Connect: &structs.ConsulConnect{ SidecarService: &structs.ConsulSidecarService{}, }, }, } t.Run("plain job", func(t *testing.T) { j := mock.Job() require.NoError(t, validateJob(j)) }) t.Run("valid consul connect", func(t *testing.T) { j := mock.Job() tg := j.TaskGroups[0] tg.Services = tgServices tg.Networks[0].Mode = "bridge" err := validateJob(j) require.NoError(t, err) }) t.Run("consul connect but missing network", func(t *testing.T) { j := mock.Job() tg := j.TaskGroups[0] tg.Services = tgServices tg.Networks = nil err := validateJob(j) require.Error(t, err) require.Contains(t, err.Error(), `Consul Connect sidecars require exactly 1 network`) }) t.Run("consul connect but non bridge network", func(t *testing.T) { j := mock.Job() tg := j.TaskGroups[0] tg.Services = tgServices tg.Networks = structs.Networks{ {Mode: "host"}, } err := validateJob(j) require.Error(t, err) require.Contains(t, err.Error(), `Consul Connect sidecar requires bridge network, found "host" in group "web"`) }) } func TestJobEndpoint_ImplicitConstraints_Signals(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request with a job asking for a template that sends a // signal job := mock.Job() signal1 := "SIGUSR1" signal2 := "SIGHUP" job.TaskGroups[0].Tasks[0].Templates = []*structs.Template{ { SourcePath: "foo", DestPath: "bar", ChangeMode: structs.TemplateChangeModeSignal, ChangeSignal: signal1, }, { SourcePath: "foo", DestPath: "baz", ChangeMode: structs.TemplateChangeModeSignal, ChangeSignal: signal2, }, } req := &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Fetch the response var resp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", req, &resp); err != nil { t.Fatalf("bad: %v", err) } // Check for the job in the FSM state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, job.ID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != resp.JobModifyIndex { t.Fatalf("index mis-match") } // Check that there is an implicit signal constraint constraints := out.TaskGroups[0].Constraints if len(constraints) != 1 { t.Fatalf("Expected an implicit constraint") } sigConstraint := getSignalConstraint([]string{signal1, signal2}) if !strings.HasPrefix(sigConstraint.RTarget, "SIGHUP") { t.Fatalf("signals not sorted: %v", sigConstraint.RTarget) } if !constraints[0].Equal(sigConstraint) { t.Fatalf("Expected implicit vault constraint") } } func TestJobEndpoint_ValidateJobUpdate(t *testing.T) { t.Parallel() require := require.New(t) old := mock.Job() new := mock.Job() if err := validateJobUpdate(old, new); err != nil { t.Errorf("expected update to be valid but got: %v", err) } new.Type = "batch" if err := validateJobUpdate(old, new); err == nil { t.Errorf("expected err when setting new job to a different type") } else { t.Log(err) } new = mock.Job() new.Periodic = &structs.PeriodicConfig{Enabled: true} if err := validateJobUpdate(old, new); err == nil { t.Errorf("expected err when setting new job to periodic") } else { t.Log(err) } new = mock.Job() new.ParameterizedJob = &structs.ParameterizedJobConfig{} if err := validateJobUpdate(old, new); err == nil { t.Errorf("expected err when setting new job to parameterized") } else { t.Log(err) } new = mock.Job() new.Dispatched = true require.Error(validateJobUpdate(old, new), "expected err when setting new job to dispatched") require.Error(validateJobUpdate(nil, new), "expected err when setting new job to dispatched") require.Error(validateJobUpdate(new, old), "expected err when setting dispatched to false") require.NoError(validateJobUpdate(nil, old)) } func TestJobEndpoint_ValidateJobUpdate_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) job := mock.Job() req := &structs.JobValidateRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Attempt to update without providing a valid token var resp structs.JobValidateResponse err := msgpackrpc.CallWithCodec(codec, "Job.Validate", req, &resp) require.NotNil(err) // Update with a valid token req.AuthToken = root.SecretID var validResp structs.JobValidateResponse err = msgpackrpc.CallWithCodec(codec, "Job.Validate", req, &validResp) require.Nil(err) require.Equal("", validResp.Error) require.Equal("", validResp.Warnings) } func TestJobEndpoint_Dispatch_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create a parameterized job job := mock.BatchJob() job.ParameterizedJob = &structs.ParameterizedJobConfig{} err := state.UpsertJob(structs.MsgTypeTestSetup, 400, job) require.Nil(err) req := &structs.JobDispatchRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Attempt to fetch the response without a token should fail var resp structs.JobDispatchResponse err = msgpackrpc.CallWithCodec(codec, "Job.Dispatch", req, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Attempt to fetch the response with an invalid token should fail invalidToken := mock.CreatePolicyAndToken(t, state, 1001, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) req.AuthToken = invalidToken.SecretID var invalidResp structs.JobDispatchResponse err = msgpackrpc.CallWithCodec(codec, "Job.Dispatch", req, &invalidResp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Dispatch with a valid management token should succeed req.AuthToken = root.SecretID var validResp structs.JobDispatchResponse err = msgpackrpc.CallWithCodec(codec, "Job.Dispatch", req, &validResp) require.Nil(err) require.NotNil(validResp.EvalID) require.NotNil(validResp.DispatchedJobID) require.NotEqual(validResp.DispatchedJobID, "") // Dispatch with a valid token should succeed validToken := mock.CreatePolicyAndToken(t, state, 1003, "test-valid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityDispatchJob})) req.AuthToken = validToken.SecretID var validResp2 structs.JobDispatchResponse err = msgpackrpc.CallWithCodec(codec, "Job.Dispatch", req, &validResp2) require.Nil(err) require.NotNil(validResp2.EvalID) require.NotNil(validResp2.DispatchedJobID) require.NotEqual(validResp2.DispatchedJobID, "") ws := memdb.NewWatchSet() out, err := state.JobByID(ws, job.Namespace, validResp2.DispatchedJobID) require.Nil(err) require.NotNil(out) require.Equal(out.ParentID, job.ID) // Look up the evaluation eval, err := state.EvalByID(ws, validResp2.EvalID) require.Nil(err) require.NotNil(eval) require.Equal(eval.CreateIndex, validResp2.EvalCreateIndex) } func TestJobEndpoint_Dispatch(t *testing.T) { t.Parallel() // No requirements d1 := mock.BatchJob() d1.ParameterizedJob = &structs.ParameterizedJobConfig{} // Require input data d2 := mock.BatchJob() d2.ParameterizedJob = &structs.ParameterizedJobConfig{ Payload: structs.DispatchPayloadRequired, } // Disallow input data d3 := mock.BatchJob() d3.ParameterizedJob = &structs.ParameterizedJobConfig{ Payload: structs.DispatchPayloadForbidden, } // Require meta d4 := mock.BatchJob() d4.ParameterizedJob = &structs.ParameterizedJobConfig{ MetaRequired: []string{"foo", "bar"}, } // Optional meta d5 := mock.BatchJob() d5.ParameterizedJob = &structs.ParameterizedJobConfig{ MetaOptional: []string{"foo", "bar"}, } // Periodic dispatch job d6 := mock.PeriodicJob() d6.ParameterizedJob = &structs.ParameterizedJobConfig{} d7 := mock.BatchJob() d7.ParameterizedJob = &structs.ParameterizedJobConfig{} d7.Stop = true reqNoInputNoMeta := &structs.JobDispatchRequest{} reqInputDataNoMeta := &structs.JobDispatchRequest{ Payload: []byte("hello world"), } reqNoInputDataMeta := &structs.JobDispatchRequest{ Meta: map[string]string{ "foo": "f1", "bar": "f2", }, } reqInputDataMeta := &structs.JobDispatchRequest{ Payload: []byte("hello world"), Meta: map[string]string{ "foo": "f1", "bar": "f2", }, } reqBadMeta := &structs.JobDispatchRequest{ Payload: []byte("hello world"), Meta: map[string]string{ "foo": "f1", "bar": "f2", "baz": "f3", }, } reqInputDataTooLarge := &structs.JobDispatchRequest{ Payload: make([]byte, DispatchPayloadSizeLimit+100), } type existingIdempotentChildJob struct { isTerminal bool } type testCase struct { name string parameterizedJob *structs.Job dispatchReq *structs.JobDispatchRequest noEval bool err bool errStr string idempotencyToken string existingIdempotentJob *existingIdempotentChildJob } cases := []testCase{ { name: "optional input data w/ data", parameterizedJob: d1, dispatchReq: reqInputDataNoMeta, err: false, }, { name: "optional input data w/o data", parameterizedJob: d1, dispatchReq: reqNoInputNoMeta, err: false, }, { name: "require input data w/ data", parameterizedJob: d2, dispatchReq: reqInputDataNoMeta, err: false, }, { name: "require input data w/o data", parameterizedJob: d2, dispatchReq: reqNoInputNoMeta, err: true, errStr: "not provided but required", }, { name: "disallow input data w/o data", parameterizedJob: d3, dispatchReq: reqNoInputNoMeta, err: false, }, { name: "disallow input data w/ data", parameterizedJob: d3, dispatchReq: reqInputDataNoMeta, err: true, errStr: "provided but forbidden", }, { name: "require meta w/ meta", parameterizedJob: d4, dispatchReq: reqInputDataMeta, err: false, }, { name: "require meta w/o meta", parameterizedJob: d4, dispatchReq: reqNoInputNoMeta, err: true, errStr: "did not provide required meta keys", }, { name: "optional meta w/ meta", parameterizedJob: d5, dispatchReq: reqNoInputDataMeta, err: false, }, { name: "optional meta w/o meta", parameterizedJob: d5, dispatchReq: reqNoInputNoMeta, err: false, }, { name: "optional meta w/ bad meta", parameterizedJob: d5, dispatchReq: reqBadMeta, err: true, errStr: "unpermitted metadata keys", }, { name: "optional input w/ too big of input", parameterizedJob: d1, dispatchReq: reqInputDataTooLarge, err: true, errStr: "Payload exceeds maximum size", }, { name: "periodic job dispatched, ensure no eval", parameterizedJob: d6, dispatchReq: reqNoInputNoMeta, noEval: true, }, { name: "periodic job stopped, ensure error", parameterizedJob: d7, dispatchReq: reqNoInputNoMeta, err: true, errStr: "stopped", }, { name: "idempotency token, no existing child job", parameterizedJob: d1, dispatchReq: reqInputDataNoMeta, err: false, idempotencyToken: "foo", existingIdempotentJob: nil, }, { name: "idempotency token, w/ existing non-terminal child job", parameterizedJob: d1, dispatchReq: reqInputDataNoMeta, err: false, idempotencyToken: "foo", existingIdempotentJob: &existingIdempotentChildJob{ isTerminal: false, }, noEval: true, }, { name: "idempotency token, w/ existing terminal job", parameterizedJob: d1, dispatchReq: reqInputDataNoMeta, err: false, idempotencyToken: "foo", existingIdempotentJob: &existingIdempotentChildJob{ isTerminal: true, }, noEval: true, }, } for _, tc := range cases { t.Run(tc.name, func(t *testing.T) { s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) // Create the register request regReq := &structs.JobRegisterRequest{ Job: tc.parameterizedJob, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: tc.parameterizedJob.Namespace, }, } // Fetch the response var regResp structs.JobRegisterResponse if err := msgpackrpc.CallWithCodec(codec, "Job.Register", regReq, ®Resp); err != nil { t.Fatalf("err: %v", err) } // Now try to dispatch tc.dispatchReq.JobID = tc.parameterizedJob.ID tc.dispatchReq.WriteRequest = structs.WriteRequest{ Region: "global", Namespace: tc.parameterizedJob.Namespace, IdempotencyToken: tc.idempotencyToken, } // Dispatch with the same request so a child job w/ the idempotency key exists var initialIdempotentDispatchResp structs.JobDispatchResponse if tc.existingIdempotentJob != nil { if err := msgpackrpc.CallWithCodec(codec, "Job.Dispatch", tc.dispatchReq, &initialIdempotentDispatchResp); err != nil { t.Fatalf("Unexpected error dispatching initial idempotent job: %v", err) } if tc.existingIdempotentJob.isTerminal { eval, err := s1.State().EvalByID(nil, initialIdempotentDispatchResp.EvalID) if err != nil { t.Fatalf("Unexpected error fetching eval %v", err) } eval = eval.Copy() eval.Status = structs.EvalStatusComplete err = s1.State().UpsertEvals(structs.MsgTypeTestSetup, initialIdempotentDispatchResp.Index+1, []*structs.Evaluation{eval}) if err != nil { t.Fatalf("Unexpected error completing eval %v", err) } } } var dispatchResp structs.JobDispatchResponse dispatchErr := msgpackrpc.CallWithCodec(codec, "Job.Dispatch", tc.dispatchReq, &dispatchResp) if dispatchErr == nil { if tc.err { t.Fatalf("Expected error: %v", dispatchErr) } // Check that we got an eval and job id back switch dispatchResp.EvalID { case "": if !tc.noEval { t.Fatalf("Bad response") } default: if tc.noEval { t.Fatalf("Got eval %q", dispatchResp.EvalID) } } if dispatchResp.DispatchedJobID == "" { t.Fatalf("Bad response") } state := s1.fsm.State() ws := memdb.NewWatchSet() out, err := state.JobByID(ws, tc.parameterizedJob.Namespace, dispatchResp.DispatchedJobID) if err != nil { t.Fatalf("err: %v", err) } if out == nil { t.Fatalf("expected job") } if out.CreateIndex != dispatchResp.JobCreateIndex { t.Fatalf("index mis-match") } if out.ParentID != tc.parameterizedJob.ID { t.Fatalf("bad parent ID") } if !out.Dispatched { t.Fatal("expected dispatched job") } if out.IsParameterized() { t.Fatal("dispatched job should not be parameterized") } if out.ParameterizedJob == nil { t.Fatal("parameter job config should exist") } // Check that the existing job is returned in the case of a supplied idempotency token if tc.idempotencyToken != "" && tc.existingIdempotentJob != nil { if dispatchResp.DispatchedJobID != initialIdempotentDispatchResp.DispatchedJobID { t.Fatal("dispatched job id should match initial dispatch") } if dispatchResp.JobCreateIndex != initialIdempotentDispatchResp.JobCreateIndex { t.Fatal("dispatched job create index should match initial dispatch") } } if tc.noEval { return } // Lookup the evaluation eval, err := state.EvalByID(ws, dispatchResp.EvalID) if err != nil { t.Fatalf("err: %v", err) } if eval == nil { t.Fatalf("expected eval") } if eval.CreateIndex != dispatchResp.EvalCreateIndex { t.Fatalf("index mis-match") } } else { if !tc.err { t.Fatalf("Got unexpected error: %v", dispatchErr) } else if !strings.Contains(dispatchErr.Error(), tc.errStr) { t.Fatalf("Expected err to include %q; got %v", tc.errStr, dispatchErr) } } }) } } // TestJobEndpoint_Dispatch_JobChildrenSummary asserts that the job summary is updated // appropriately as its dispatched/children jobs status are updated. func TestJobEndpoint_Dispatch_JobChildrenSummary(t *testing.T) { t.Parallel() s1, cleanupS1 := TestServer(t, func(c *Config) { c.NumSchedulers = 0 // Prevent automatic dequeue }) defer cleanupS1() state := s1.fsm.State() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) node := mock.Node() require.NoError(t, state.UpsertNode(structs.MsgTypeTestSetup, 1, node)) parameterizedJob := mock.BatchJob() parameterizedJob.ParameterizedJob = &structs.ParameterizedJobConfig{} // Create the register request regReq := &structs.JobRegisterRequest{ Job: parameterizedJob, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: parameterizedJob.Namespace, }, } var regResp structs.JobRegisterResponse require.NoError(t, msgpackrpc.CallWithCodec(codec, "Job.Register", regReq, ®Resp)) jobChildren := func() *structs.JobChildrenSummary { summary, err := state.JobSummaryByID(nil, parameterizedJob.Namespace, parameterizedJob.ID) require.NoError(t, err) return summary.Children } require.Equal(t, &structs.JobChildrenSummary{}, jobChildren()) // dispatch a child job dispatchReq := &structs.JobDispatchRequest{ JobID: parameterizedJob.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: parameterizedJob.Namespace, }, } var dispatchResp structs.JobDispatchResponse err := msgpackrpc.CallWithCodec(codec, "Job.Dispatch", dispatchReq, &dispatchResp) require.NoError(t, err) nextIdx := dispatchResp.Index + 1 require.Equal(t, &structs.JobChildrenSummary{Pending: 1}, jobChildren()) dispatchedJob, err := state.JobByID(nil, parameterizedJob.Namespace, dispatchResp.DispatchedJobID) require.NoError(t, err) require.NotNil(t, dispatchedJob) dispatchedStatus := func() string { job, err := state.JobByID(nil, dispatchedJob.Namespace, dispatchedJob.ID) require.NoError(t, err) require.NotNil(t, job) return job.Status } // Let's start an alloc for the dispatch job and walk through states // Note that job summary reports 1 running even when alloc is pending! nextIdx++ alloc := mock.Alloc() alloc.Job = dispatchedJob alloc.JobID = dispatchedJob.ID alloc.TaskGroup = dispatchedJob.TaskGroups[0].Name alloc.Namespace = dispatchedJob.Namespace alloc.ClientStatus = structs.AllocClientStatusPending err = s1.State().UpsertAllocs(structs.MsgTypeTestSetup, nextIdx, []*structs.Allocation{alloc}) require.NoError(t, err) require.Equal(t, &structs.JobChildrenSummary{Running: 1}, jobChildren()) require.Equal(t, structs.JobStatusRunning, dispatchedStatus()) // mark the creation eval as completed nextIdx++ eval, err := state.EvalByID(nil, dispatchResp.EvalID) require.NoError(t, err) eval = eval.Copy() eval.Status = structs.EvalStatusComplete require.NoError(t, state.UpsertEvals(structs.MsgTypeTestSetup, nextIdx, []*structs.Evaluation{eval})) updateAllocStatus := func(status string) { nextIdx++ nalloc, err := state.AllocByID(nil, alloc.ID) require.NoError(t, err) nalloc = nalloc.Copy() nalloc.ClientStatus = status err = s1.State().UpdateAllocsFromClient(structs.MsgTypeTestSetup, nextIdx, []*structs.Allocation{nalloc}) require.NoError(t, err) } // job should remain remaining when alloc runs updateAllocStatus(structs.AllocClientStatusRunning) require.Equal(t, &structs.JobChildrenSummary{Running: 1}, jobChildren()) require.Equal(t, structs.JobStatusRunning, dispatchedStatus()) // job should be dead after alloc completes updateAllocStatus(structs.AllocClientStatusComplete) require.Equal(t, &structs.JobChildrenSummary{Dead: 1}, jobChildren()) require.Equal(t, structs.JobStatusDead, dispatchedStatus()) } func TestJobEndpoint_Dispatch_ACL_RejectedBySchedulerConfig(t *testing.T) { t.Parallel() s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job := mock.BatchJob() job.ParameterizedJob = &structs.ParameterizedJobConfig{} err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.NoError(t, err) dispatch := &structs.JobDispatchRequest{ JobID: job.ID, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } submitJobToken := mock.CreatePolicyAndToken(t, state, 1001, "test-valid-write", mock.NamespacePolicy(structs.DefaultNamespace, "write", nil)). SecretID cases := []struct { name string token string rejectEnabled bool errExpected string }{ { name: "reject disabled, with a submit token", token: submitJobToken, rejectEnabled: false, }, { name: "reject enabled, with a submit token", token: submitJobToken, rejectEnabled: true, errExpected: structs.ErrJobRegistrationDisabled.Error(), }, { name: "reject enabled, without a token", token: "", rejectEnabled: true, errExpected: structs.ErrPermissionDenied.Error(), }, { name: "reject enabled, with a management token", token: root.SecretID, rejectEnabled: true, }, } for _, tc := range cases { t.Run(tc.name, func(t *testing.T) { cfgReq := &structs.SchedulerSetConfigRequest{ Config: structs.SchedulerConfiguration{ RejectJobRegistration: tc.rejectEnabled, }, WriteRequest: structs.WriteRequest{ Region: "global", }, } cfgReq.AuthToken = root.SecretID err := msgpackrpc.CallWithCodec(codec, "Operator.SchedulerSetConfiguration", cfgReq, &structs.SchedulerSetConfigurationResponse{}, ) require.NoError(t, err) dispatch.AuthToken = tc.token var resp structs.JobDispatchResponse err = msgpackrpc.CallWithCodec(codec, "Job.Dispatch", dispatch, &resp) if tc.errExpected != "" { require.Error(t, err, "expected error") require.EqualError(t, err, tc.errExpected) } else { require.NoError(t, err, "unexpected error") require.NotEqual(t, 0, resp.Index) } }) } } func TestJobEndpoint_Scale(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job := mock.Job() originalCount := job.TaskGroups[0].Count err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) groupName := job.TaskGroups[0].Name scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: groupName, }, Count: helper.Int64ToPtr(int64(originalCount + 1)), Message: "because of the load", Meta: map[string]interface{}{ "metrics": map[string]string{ "1": "a", "2": "b", }, "other": "value", }, PolicyOverride: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.NoError(err) require.NotEmpty(resp.EvalID) require.Greater(resp.EvalCreateIndex, resp.JobModifyIndex) events, _, _ := state.ScalingEventsByJob(nil, job.Namespace, job.ID) require.Equal(1, len(events[groupName])) require.Equal(int64(originalCount), events[groupName][0].PreviousCount) } func TestJobEndpoint_Scale_DeploymentBlocking(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() type testCase struct { latestDeploymentStatus string } cases := []string{ structs.DeploymentStatusSuccessful, structs.DeploymentStatusPaused, structs.DeploymentStatusRunning, } for _, tc := range cases { // create a job with a deployment history job := mock.Job() require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 1000, job), "UpsertJob") d1 := mock.Deployment() d1.Status = structs.DeploymentStatusCancelled d1.StatusDescription = structs.DeploymentStatusDescriptionNewerJob d1.JobID = job.ID d1.JobCreateIndex = job.CreateIndex require.Nil(state.UpsertDeployment(1001, d1), "UpsertDeployment") d2 := mock.Deployment() d2.Status = structs.DeploymentStatusSuccessful d2.StatusDescription = structs.DeploymentStatusDescriptionSuccessful d2.JobID = job.ID d2.JobCreateIndex = job.CreateIndex require.Nil(state.UpsertDeployment(1002, d2), "UpsertDeployment") // add the latest deployment for the test case dLatest := mock.Deployment() dLatest.Status = tc dLatest.StatusDescription = "description does not matter for this test" dLatest.JobID = job.ID dLatest.JobCreateIndex = job.CreateIndex require.Nil(state.UpsertDeployment(1003, dLatest), "UpsertDeployment") // attempt to scale originalCount := job.TaskGroups[0].Count newCount := int64(originalCount + 1) groupName := job.TaskGroups[0].Name scalingMetadata := map[string]interface{}{ "meta": "data", } scalingMessage := "original reason for scaling" scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: groupName, }, Meta: scalingMetadata, Message: scalingMessage, Count: helper.Int64ToPtr(newCount), WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) if dLatest.Active() { // should fail require.Error(err, "test case %q", tc) require.Contains(err.Error(), "active deployment") } else { require.NoError(err, "test case %q", tc) require.NotEmpty(resp.EvalID) require.Greater(resp.EvalCreateIndex, resp.JobModifyIndex) } } } func TestJobEndpoint_Scale_InformationalEventsShouldNotBeBlocked(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() type testCase struct { latestDeploymentStatus string } cases := []string{ structs.DeploymentStatusSuccessful, structs.DeploymentStatusPaused, structs.DeploymentStatusRunning, } for _, tc := range cases { // create a job with a deployment history job := mock.Job() require.Nil(state.UpsertJob(structs.MsgTypeTestSetup, 1000, job), "UpsertJob") d1 := mock.Deployment() d1.Status = structs.DeploymentStatusCancelled d1.StatusDescription = structs.DeploymentStatusDescriptionNewerJob d1.JobID = job.ID d1.JobCreateIndex = job.CreateIndex require.Nil(state.UpsertDeployment(1001, d1), "UpsertDeployment") d2 := mock.Deployment() d2.Status = structs.DeploymentStatusSuccessful d2.StatusDescription = structs.DeploymentStatusDescriptionSuccessful d2.JobID = job.ID d2.JobCreateIndex = job.CreateIndex require.Nil(state.UpsertDeployment(1002, d2), "UpsertDeployment") // add the latest deployment for the test case dLatest := mock.Deployment() dLatest.Status = tc dLatest.StatusDescription = "description does not matter for this test" dLatest.JobID = job.ID dLatest.JobCreateIndex = job.CreateIndex require.Nil(state.UpsertDeployment(1003, dLatest), "UpsertDeployment") // register informational scaling event groupName := job.TaskGroups[0].Name scalingMetadata := map[string]interface{}{ "meta": "data", } scalingMessage := "original reason for scaling" scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: groupName, }, Meta: scalingMetadata, Message: scalingMessage, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.NoError(err, "test case %q", tc) require.Empty(resp.EvalID) events, _, _ := state.ScalingEventsByJob(nil, job.Namespace, job.ID) require.Equal(1, len(events[groupName])) latestEvent := events[groupName][0] require.False(latestEvent.Error) require.Nil(latestEvent.Count) require.Equal(scalingMessage, latestEvent.Message) require.Equal(scalingMetadata, latestEvent.Meta) } } func TestJobEndpoint_Scale_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: job.TaskGroups[0].Name, }, Message: "because of the load", WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } // Scale without a token should fail var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) scale.AuthToken = invalidToken.SecretID var invalidResp structs.JobRegisterResponse require.NotNil(err) err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &invalidResp) require.Contains(err.Error(), "Permission denied") type testCase struct { authToken string name string } cases := []testCase{ { name: "mgmt token should succeed", authToken: root.SecretID, }, { name: "write disposition should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-write", mock.NamespacePolicy(structs.DefaultNamespace, "write", nil)). SecretID, }, { name: "autoscaler disposition should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-autoscaler", mock.NamespacePolicy(structs.DefaultNamespace, "scale", nil)). SecretID, }, { name: "submit-job capability should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-submit-job", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilitySubmitJob})).SecretID, }, { name: "scale-job capability should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-scale-job", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityScaleJob})). SecretID, }, } for _, tc := range cases { scale.AuthToken = tc.authToken var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.NoError(err, tc.name) require.NotNil(resp.EvalID) } } func TestJobEndpoint_Scale_ACL_RejectedBySchedulerConfig(t *testing.T) { t.Parallel() s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.NoError(t, err) scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: job.TaskGroups[0].Name, }, Message: "because of the load", WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } submitJobToken := mock.CreatePolicyAndToken(t, state, 1001, "test-valid-write", mock.NamespacePolicy(structs.DefaultNamespace, "write", nil)). SecretID cases := []struct { name string token string rejectEnabled bool errExpected string }{ { name: "reject disabled, with a submit token", token: submitJobToken, rejectEnabled: false, }, { name: "reject enabled, with a submit token", token: submitJobToken, rejectEnabled: true, errExpected: structs.ErrJobRegistrationDisabled.Error(), }, { name: "reject enabled, without a token", token: "", rejectEnabled: true, errExpected: structs.ErrPermissionDenied.Error(), }, { name: "reject enabled, with a management token", token: root.SecretID, rejectEnabled: true, }, } for _, tc := range cases { t.Run(tc.name, func(t *testing.T) { cfgReq := &structs.SchedulerSetConfigRequest{ Config: structs.SchedulerConfiguration{ RejectJobRegistration: tc.rejectEnabled, }, WriteRequest: structs.WriteRequest{ Region: "global", }, } cfgReq.AuthToken = root.SecretID err := msgpackrpc.CallWithCodec(codec, "Operator.SchedulerSetConfiguration", cfgReq, &structs.SchedulerSetConfigurationResponse{}, ) require.NoError(t, err) var resp structs.JobRegisterResponse scale.AuthToken = tc.token err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) if tc.errExpected != "" { require.Error(t, err, "expected error") require.EqualError(t, err, tc.errExpected) } else { require.NoError(t, err, "unexpected error") require.NotEqual(t, 0, resp.Index) } }) } } func TestJobEndpoint_Scale_Invalid(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job := mock.Job() count := job.TaskGroups[0].Count // check before job registration scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: job.TaskGroups[0].Name, }, Count: helper.Int64ToPtr(int64(count) + 1), Message: "this should fail", Meta: map[string]interface{}{ "metrics": map[string]string{ "1": "a", "2": "b", }, "other": "value", }, PolicyOverride: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.Error(err) require.Contains(err.Error(), "not found") // register the job err = state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) scale.Count = helper.Int64ToPtr(10) scale.Message = "error message" scale.Error = true err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.Error(err) require.Contains(err.Error(), "should not contain count if error is true") } func TestJobEndpoint_Scale_OutOfBounds(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job, pol := mock.JobWithScalingPolicy() pol.Min = 3 pol.Max = 10 job.TaskGroups[0].Count = 5 // register the job err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) var resp structs.JobRegisterResponse scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: job.TaskGroups[0].Name, }, Count: helper.Int64ToPtr(pol.Max + 1), Message: "out of bounds", PolicyOverride: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.Error(err) require.Contains(err.Error(), "group count was greater than scaling policy maximum: 11 > 10") scale.Count = helper.Int64ToPtr(2) err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.Error(err) require.Contains(err.Error(), "group count was less than scaling policy minimum: 2 < 3") } func TestJobEndpoint_Scale_NoEval(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job := mock.Job() groupName := job.TaskGroups[0].Name originalCount := job.TaskGroups[0].Count var resp structs.JobRegisterResponse err := msgpackrpc.CallWithCodec(codec, "Job.Register", &structs.JobRegisterRequest{ Job: job, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, }, &resp) jobCreateIndex := resp.Index require.NoError(err) scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: groupName, }, Count: nil, // no count => no eval Message: "something informative", Meta: map[string]interface{}{ "metrics": map[string]string{ "1": "a", "2": "b", }, "other": "value", }, PolicyOverride: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.NoError(err) require.Empty(resp.EvalID) require.Empty(resp.EvalCreateIndex) jobEvents, eventsIndex, err := state.ScalingEventsByJob(nil, job.Namespace, job.ID) require.NoError(err) require.NotNil(jobEvents) require.Len(jobEvents, 1) require.Contains(jobEvents, groupName) groupEvents := jobEvents[groupName] require.Len(groupEvents, 1) event := groupEvents[0] require.Nil(event.EvalID) require.Greater(eventsIndex, jobCreateIndex) events, _, _ := state.ScalingEventsByJob(nil, job.Namespace, job.ID) require.Equal(1, len(events[groupName])) require.Equal(int64(originalCount), events[groupName][0].PreviousCount) } func TestJobEndpoint_Scale_Priority(t *testing.T) { t.Parallel() requireAssertion := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) fsmState := s1.fsm.State() // Create a job and alter the priority. job := mock.Job() job.Priority = 90 originalCount := job.TaskGroups[0].Count err := fsmState.UpsertJob(structs.MsgTypeTestSetup, 1000, job) requireAssertion.Nil(err) groupName := job.TaskGroups[0].Name scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: groupName, }, Count: helper.Int64ToPtr(int64(originalCount + 1)), Message: "scotty, we need more power", PolicyOverride: false, WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) requireAssertion.NoError(err) requireAssertion.NotEmpty(resp.EvalID) requireAssertion.Greater(resp.EvalCreateIndex, resp.JobModifyIndex) // Check the evaluation priority matches the job priority. eval, err := fsmState.EvalByID(nil, resp.EvalID) requireAssertion.Nil(err) requireAssertion.NotNil(eval) requireAssertion.EqualValues(resp.EvalCreateIndex, eval.CreateIndex) requireAssertion.Equal(job.Priority, eval.Priority) requireAssertion.Equal(job.Type, eval.Type) requireAssertion.Equal(structs.EvalTriggerScaling, eval.TriggeredBy) requireAssertion.Equal(job.ID, eval.JobID) requireAssertion.NotZero(eval.CreateTime) requireAssertion.NotZero(eval.ModifyTime) } func TestJobEndpoint_InvalidCount(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) scale := &structs.JobScaleRequest{ JobID: job.ID, Target: map[string]string{ structs.ScalingTargetGroup: job.TaskGroups[0].Name, }, Count: helper.Int64ToPtr(int64(-1)), WriteRequest: structs.WriteRequest{ Region: "global", Namespace: job.Namespace, }, } var resp structs.JobRegisterResponse err = msgpackrpc.CallWithCodec(codec, "Job.Scale", scale, &resp) require.Error(err) } func TestJobEndpoint_GetScaleStatus(t *testing.T) { t.Parallel() require := require.New(t) s1, cleanupS1 := TestServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() jobV1 := mock.Job() // check before registration // Fetch the scaling status get := &structs.JobScaleStatusRequest{ JobID: jobV1.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: jobV1.Namespace, }, } var resp2 structs.JobScaleStatusResponse require.NoError(msgpackrpc.CallWithCodec(codec, "Job.ScaleStatus", get, &resp2)) require.Nil(resp2.JobScaleStatus) // stopped (previous version) require.NoError(state.UpsertJob(structs.MsgTypeTestSetup, 1000, jobV1), "UpsertJob") a0 := mock.Alloc() a0.Job = jobV1 a0.Namespace = jobV1.Namespace a0.JobID = jobV1.ID a0.ClientStatus = structs.AllocClientStatusComplete require.NoError(state.UpsertAllocs(structs.MsgTypeTestSetup, 1010, []*structs.Allocation{a0}), "UpsertAllocs") jobV2 := jobV1.Copy() require.NoError(state.UpsertJob(structs.MsgTypeTestSetup, 1100, jobV2), "UpsertJob") a1 := mock.Alloc() a1.Job = jobV2 a1.Namespace = jobV2.Namespace a1.JobID = jobV2.ID a1.ClientStatus = structs.AllocClientStatusRunning // healthy a1.DeploymentStatus = &structs.AllocDeploymentStatus{ Healthy: helper.BoolToPtr(true), } a2 := mock.Alloc() a2.Job = jobV2 a2.Namespace = jobV2.Namespace a2.JobID = jobV2.ID a2.ClientStatus = structs.AllocClientStatusPending // unhealthy a2.DeploymentStatus = &structs.AllocDeploymentStatus{ Healthy: helper.BoolToPtr(false), } a3 := mock.Alloc() a3.Job = jobV2 a3.Namespace = jobV2.Namespace a3.JobID = jobV2.ID a3.ClientStatus = structs.AllocClientStatusRunning // canary a3.DeploymentStatus = &structs.AllocDeploymentStatus{ Healthy: helper.BoolToPtr(true), Canary: true, } // no health a4 := mock.Alloc() a4.Job = jobV2 a4.Namespace = jobV2.Namespace a4.JobID = jobV2.ID a4.ClientStatus = structs.AllocClientStatusRunning // upsert allocations require.NoError(state.UpsertAllocs(structs.MsgTypeTestSetup, 1110, []*structs.Allocation{a1, a2, a3, a4}), "UpsertAllocs") event := &structs.ScalingEvent{ Time: time.Now().Unix(), Count: helper.Int64ToPtr(5), Message: "message", Error: false, Meta: map[string]interface{}{ "a": "b", }, EvalID: nil, } require.NoError(state.UpsertScalingEvent(1003, &structs.ScalingEventRequest{ Namespace: jobV2.Namespace, JobID: jobV2.ID, TaskGroup: jobV2.TaskGroups[0].Name, ScalingEvent: event, }), "UpsertScalingEvent") // check after job registration require.NoError(msgpackrpc.CallWithCodec(codec, "Job.ScaleStatus", get, &resp2)) require.NotNil(resp2.JobScaleStatus) expectedStatus := structs.JobScaleStatus{ JobID: jobV2.ID, Namespace: jobV2.Namespace, JobCreateIndex: jobV2.CreateIndex, JobModifyIndex: a1.CreateIndex, JobStopped: jobV2.Stop, TaskGroups: map[string]*structs.TaskGroupScaleStatus{ jobV2.TaskGroups[0].Name: { Desired: jobV2.TaskGroups[0].Count, Placed: 3, Running: 2, Healthy: 1, Unhealthy: 1, Events: []*structs.ScalingEvent{event}, }, }, } require.True(reflect.DeepEqual(*resp2.JobScaleStatus, expectedStatus)) } func TestJobEndpoint_GetScaleStatus_ACL(t *testing.T) { t.Parallel() require := require.New(t) s1, root, cleanupS1 := TestACLServer(t, nil) defer cleanupS1() codec := rpcClient(t, s1) testutil.WaitForLeader(t, s1.RPC) state := s1.fsm.State() // Create the job job := mock.Job() err := state.UpsertJob(structs.MsgTypeTestSetup, 1000, job) require.Nil(err) // Get the job scale status get := &structs.JobScaleStatusRequest{ JobID: job.ID, QueryOptions: structs.QueryOptions{ Region: "global", Namespace: job.Namespace, }, } // Get without a token should fail var resp structs.JobScaleStatusResponse err = msgpackrpc.CallWithCodec(codec, "Job.ScaleStatus", get, &resp) require.NotNil(err) require.Contains(err.Error(), "Permission denied") // Expect failure for request with an invalid token invalidToken := mock.CreatePolicyAndToken(t, state, 1003, "test-invalid", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityListJobs})) get.AuthToken = invalidToken.SecretID require.NotNil(err) err = msgpackrpc.CallWithCodec(codec, "Job.ScaleStatus", get, &resp) require.Contains(err.Error(), "Permission denied") type testCase struct { authToken string name string } cases := []testCase{ { name: "mgmt token should succeed", authToken: root.SecretID, }, { name: "read disposition should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-read", mock.NamespacePolicy(structs.DefaultNamespace, "read", nil)). SecretID, }, { name: "write disposition should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-write", mock.NamespacePolicy(structs.DefaultNamespace, "write", nil)). SecretID, }, { name: "autoscaler disposition should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-autoscaler", mock.NamespacePolicy(structs.DefaultNamespace, "scale", nil)). SecretID, }, { name: "read-job capability should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-read-job", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})).SecretID, }, { name: "read-job-scaling capability should succeed", authToken: mock.CreatePolicyAndToken(t, state, 1005, "test-valid-read-job-scaling", mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJobScaling})). SecretID, }, } for _, tc := range cases { get.AuthToken = tc.authToken var validResp structs.JobScaleStatusResponse err = msgpackrpc.CallWithCodec(codec, "Job.ScaleStatus", get, &validResp) require.NoError(err, tc.name) require.NotNil(validResp.JobScaleStatus) } }