open-nomad/nomad/client_alloc_endpoint_test.go

1252 lines
34 KiB
Go
Raw Normal View History

2018-02-06 01:20:42 +00:00
package nomad
import (
"encoding/json"
2018-02-06 01:20:42 +00:00
"fmt"
"io"
"net"
"strings"
2018-02-06 01:20:42 +00:00
"testing"
"time"
2018-02-06 01:20:42 +00:00
msgpackrpc "github.com/hashicorp/net-rpc-msgpackrpc"
"github.com/hashicorp/nomad/acl"
"github.com/hashicorp/nomad/client"
"github.com/hashicorp/nomad/client/config"
cstructs "github.com/hashicorp/nomad/client/structs"
"github.com/hashicorp/nomad/helper/uuid"
"github.com/hashicorp/nomad/nomad/mock"
"github.com/hashicorp/nomad/nomad/structs"
nstructs "github.com/hashicorp/nomad/nomad/structs"
"github.com/hashicorp/nomad/plugins/drivers"
2018-02-06 01:20:42 +00:00
"github.com/hashicorp/nomad/testutil"
"github.com/kr/pretty"
2018-02-06 01:20:42 +00:00
"github.com/stretchr/testify/require"
"github.com/ugorji/go/codec"
2018-02-06 01:20:42 +00:00
)
func TestClientAllocations_GarbageCollectAll_Local(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
2018-02-06 01:20:42 +00:00
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
c, cleanupC := client.TestClient(t, func(c *config.Config) {
2018-02-06 01:20:42 +00:00
c.Servers = []string{s.config.RPCAddr.String()}
})
defer cleanupC()
2018-02-06 01:20:42 +00:00
testutil.WaitForResult(func() (bool, error) {
nodes := s.connectedNodes()
return len(nodes) == 1, nil
}, func(err error) {
t.Fatalf("should have a clients")
})
// Make the request without having a node-id
req := &structs.NodeSpecificRequest{
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollectAll", req, &resp)
require.NotNil(err)
require.Contains(err.Error(), "missing")
// Fetch the response setting the node id
req.NodeID = c.NodeID()
var resp2 structs.GenericResponse
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollectAll", req, &resp2)
require.Nil(err)
}
func TestClientAllocations_GarbageCollectAll_Local_ACL(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server
s, root, cleanupS := TestACLServer(t, nil)
defer cleanupS()
2018-02-06 01:20:42 +00:00
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Create a bad token
policyBad := mock.NamespacePolicy("other", "", []string{acl.NamespaceCapabilityReadFS})
tokenBad := mock.CreatePolicyAndToken(t, s.State(), 1005, "invalid", policyBad)
policyGood := mock.NodePolicy(acl.PolicyWrite)
tokenGood := mock.CreatePolicyAndToken(t, s.State(), 1009, "valid2", policyGood)
cases := []struct {
Name string
Token string
ExpectedError string
}{
{
Name: "bad token",
Token: tokenBad.SecretID,
ExpectedError: structs.ErrPermissionDenied.Error(),
},
{
Name: "good token",
Token: tokenGood.SecretID,
ExpectedError: "Unknown node",
},
{
Name: "root token",
Token: root.SecretID,
ExpectedError: "Unknown node",
},
}
for _, c := range cases {
t.Run(c.Name, func(t *testing.T) {
// Make the request without having a node-id
req := &structs.NodeSpecificRequest{
NodeID: uuid.Generate(),
QueryOptions: structs.QueryOptions{
AuthToken: c.Token,
Region: "global",
},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollectAll", req, &resp)
require.NotNil(err)
require.Contains(err.Error(), c.ExpectedError)
})
}
}
func TestClientAllocations_GarbageCollectAll_NoNode(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
2018-02-06 01:20:42 +00:00
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Make the request without having a node-id
req := &structs.NodeSpecificRequest{
NodeID: uuid.Generate(),
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollectAll", req, &resp)
require.NotNil(err)
require.Contains(err.Error(), "Unknown node")
}
func TestClientAllocations_GarbageCollectAll_OldNode(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and fake an old client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
state := s.State()
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Test for an old version error
node := mock.Node()
node.Attributes["nomad.version"] = "0.7.1"
require.Nil(state.UpsertNode(1005, node))
req := &structs.NodeSpecificRequest{
NodeID: node.ID,
QueryOptions: structs.QueryOptions{Region: "global"},
}
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollectAll", req, &resp)
require.True(structs.IsErrNodeLacksRpc(err))
// Test for a missing version error
delete(node.Attributes, "nomad.version")
require.Nil(state.UpsertNode(1006, node))
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollectAll", req, &resp)
require.True(structs.IsErrUnknownNomadVersion(err))
}
2018-02-06 01:20:42 +00:00
func TestClientAllocations_GarbageCollectAll_Remote(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
s1, cleanupS1 := TestServer(t, func(c *Config) {
c.BootstrapExpect = 2
})
defer cleanupS1()
s2, cleanupS2 := TestServer(t, func(c *Config) {
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
c.BootstrapExpect = 2
2018-02-06 01:20:42 +00:00
})
defer cleanupS2()
2018-02-06 01:20:42 +00:00
TestJoin(t, s1, s2)
testutil.WaitForLeader(t, s1.RPC)
testutil.WaitForLeader(t, s2.RPC)
codec := rpcClient(t, s2)
c, cleanupC := client.TestClient(t, func(c *config.Config) {
2018-02-06 01:20:42 +00:00
c.Servers = []string{s2.config.RPCAddr.String()}
2018-02-14 00:31:17 +00:00
c.GCDiskUsageThreshold = 100.0
2018-02-06 01:20:42 +00:00
})
defer cleanupC()
2018-02-06 01:20:42 +00:00
testutil.WaitForResult(func() (bool, error) {
nodes := s2.connectedNodes()
if len(nodes) != 1 {
return false, fmt.Errorf("should have 1 client. found %d", len(nodes))
}
req := &structs.NodeSpecificRequest{
NodeID: c.NodeID(),
QueryOptions: structs.QueryOptions{Region: "global"},
}
resp := structs.SingleNodeResponse{}
if err := msgpackrpc.CallWithCodec(codec, "Node.GetNode", req, &resp); err != nil {
return false, err
}
return resp.Node != nil && resp.Node.Status == structs.NodeStatusReady, fmt.Errorf(
"expected ready but found %s", pretty.Sprint(resp.Node))
2018-02-06 01:20:42 +00:00
}, func(err error) {
t.Fatalf("should have a clients")
})
// Force remove the connection locally in case it exists
s1.nodeConnsLock.Lock()
delete(s1.nodeConns, c.NodeID())
s1.nodeConnsLock.Unlock()
// Make the request
req := &structs.NodeSpecificRequest{
NodeID: c.NodeID(),
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp cstructs.ClientStatsResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollectAll", req, &resp)
require.Nil(err)
}
func TestClientAllocations_GarbageCollect_OldNode(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and fake an old client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
state := s.State()
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Test for an old version error
node := mock.Node()
node.Attributes["nomad.version"] = "0.7.1"
require.Nil(state.UpsertNode(1005, node))
alloc := mock.Alloc()
alloc.NodeID = node.ID
require.Nil(state.UpsertAllocs(1006, []*structs.Allocation{alloc}))
req := &structs.AllocSpecificRequest{
AllocID: alloc.ID,
QueryOptions: structs.QueryOptions{
Region: "global",
Namespace: structs.DefaultNamespace,
},
}
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollect", req, &resp)
require.True(structs.IsErrNodeLacksRpc(err), err.Error())
// Test for a missing version error
delete(node.Attributes, "nomad.version")
require.Nil(state.UpsertNode(1007, node))
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollect", req, &resp)
require.True(structs.IsErrUnknownNomadVersion(err), err.Error())
}
2018-02-06 01:20:42 +00:00
func TestClientAllocations_GarbageCollect_Local(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
2018-02-06 01:20:42 +00:00
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
c, cleanupC := client.TestClient(t, func(c *config.Config) {
2018-02-06 01:20:42 +00:00
c.Servers = []string{s.config.RPCAddr.String()}
2018-02-14 00:31:17 +00:00
c.GCDiskUsageThreshold = 100.0
2018-02-06 01:20:42 +00:00
})
defer cleanupC()
2018-02-06 01:20:42 +00:00
// Force an allocation onto the node
a := mock.Alloc()
a.Job.Type = structs.JobTypeBatch
a.NodeID = c.NodeID()
a.Job.TaskGroups[0].Count = 1
a.Job.TaskGroups[0].Tasks[0] = &structs.Task{
Name: "web",
Driver: "mock_driver",
Config: map[string]interface{}{
"run_for": "2s",
2018-02-06 01:20:42 +00:00
},
LogConfig: structs.DefaultLogConfig(),
Resources: &structs.Resources{
CPU: 500,
MemoryMB: 256,
},
}
testutil.WaitForResult(func() (bool, error) {
nodes := s.connectedNodes()
return len(nodes) == 1, nil
}, func(err error) {
t.Fatalf("should have a clients")
})
// Upsert the allocation
state := s.State()
require.Nil(state.UpsertJob(999, a.Job))
require.Nil(state.UpsertAllocs(1003, []*structs.Allocation{a}))
// Wait for the client to run the allocation
testutil.WaitForResult(func() (bool, error) {
alloc, err := state.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
if alloc.ClientStatus != structs.AllocClientStatusComplete {
return false, fmt.Errorf("alloc client status: %v", alloc.ClientStatus)
}
return true, nil
}, func(err error) {
t.Fatalf("Alloc on node %q not finished: %v", c.NodeID(), err)
})
// Make the request without having an alloc id
req := &structs.AllocSpecificRequest{
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollect", req, &resp)
require.NotNil(err)
require.Contains(err.Error(), "missing")
// Fetch the response setting the node id
req.AllocID = a.ID
var resp2 structs.GenericResponse
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollect", req, &resp2)
require.Nil(err)
}
func TestClientAllocations_GarbageCollect_Local_ACL(t *testing.T) {
t.Parallel()
// Start a server
s, root, cleanupS := TestACLServer(t, nil)
defer cleanupS()
2018-02-06 01:20:42 +00:00
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Create a bad token
policyBad := mock.NamespacePolicy("other", "", []string{acl.NamespaceCapabilityReadFS})
tokenBad := mock.CreatePolicyAndToken(t, s.State(), 1005, "invalid", policyBad)
policyGood := mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilitySubmitJob})
tokenGood := mock.CreatePolicyAndToken(t, s.State(), 1009, "valid2", policyGood)
// Upsert the allocation
state := s.State()
alloc := mock.Alloc()
require.NoError(t, state.UpsertJob(1010, alloc.Job))
require.NoError(t, state.UpsertAllocs(1011, []*structs.Allocation{alloc}))
2018-02-06 01:20:42 +00:00
cases := []struct {
Name string
Token string
ExpectedError string
}{
{
Name: "bad token",
Token: tokenBad.SecretID,
ExpectedError: structs.ErrPermissionDenied.Error(),
},
{
Name: "good token",
Token: tokenGood.SecretID,
ExpectedError: structs.ErrUnknownNodePrefix,
2018-02-06 01:20:42 +00:00
},
{
Name: "root token",
Token: root.SecretID,
ExpectedError: structs.ErrUnknownNodePrefix,
2018-02-06 01:20:42 +00:00
},
}
for _, c := range cases {
t.Run(c.Name, func(t *testing.T) {
// Make the request without having a node-id
req := &structs.AllocSpecificRequest{
AllocID: alloc.ID,
2018-02-06 01:20:42 +00:00
QueryOptions: structs.QueryOptions{
AuthToken: c.Token,
Region: "global",
Namespace: structs.DefaultNamespace,
},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollect", req, &resp)
require.NotNil(t, err)
require.Contains(t, err.Error(), c.ExpectedError)
2018-02-06 01:20:42 +00:00
})
}
}
func TestClientAllocations_GarbageCollect_Remote(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
s1, cleanupS1 := TestServer(t, func(c *Config) {
c.BootstrapExpect = 2
})
defer cleanupS1()
s2, cleanupS2 := TestServer(t, func(c *Config) {
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
c.BootstrapExpect = 2
2018-02-06 01:20:42 +00:00
})
defer cleanupS2()
2018-02-06 01:20:42 +00:00
TestJoin(t, s1, s2)
testutil.WaitForLeader(t, s1.RPC)
testutil.WaitForLeader(t, s2.RPC)
codec := rpcClient(t, s2)
c, cleanup := client.TestClient(t, func(c *config.Config) {
2018-02-06 01:20:42 +00:00
c.Servers = []string{s2.config.RPCAddr.String()}
2018-02-14 00:31:17 +00:00
c.GCDiskUsageThreshold = 100.0
2018-02-06 01:20:42 +00:00
})
defer cleanup()
2018-02-06 01:20:42 +00:00
// Force an allocation onto the node
a := mock.Alloc()
a.Job.Type = structs.JobTypeBatch
a.NodeID = c.NodeID()
a.Job.TaskGroups[0].Count = 1
a.Job.TaskGroups[0].Tasks[0] = &structs.Task{
Name: "web",
Driver: "mock_driver",
Config: map[string]interface{}{
"run_for": "2s",
2018-02-06 01:20:42 +00:00
},
LogConfig: structs.DefaultLogConfig(),
Resources: &structs.Resources{
CPU: 500,
MemoryMB: 256,
},
}
testutil.WaitForResult(func() (bool, error) {
nodes := s2.connectedNodes()
if len(nodes) != 1 {
return false, fmt.Errorf("should have 1 client. found %d", len(nodes))
}
req := &structs.NodeSpecificRequest{
NodeID: c.NodeID(),
QueryOptions: structs.QueryOptions{Region: "global"},
}
resp := structs.SingleNodeResponse{}
if err := msgpackrpc.CallWithCodec(codec, "Node.GetNode", req, &resp); err != nil {
return false, err
}
return resp.Node != nil && resp.Node.Status == structs.NodeStatusReady, fmt.Errorf(
"expected ready but found %s", pretty.Sprint(resp.Node))
2018-02-06 01:20:42 +00:00
}, func(err error) {
t.Fatalf("should have a clients")
})
// Upsert the allocation
state1 := s1.State()
state2 := s2.State()
require.Nil(state1.UpsertJob(999, a.Job))
require.Nil(state1.UpsertAllocs(1003, []*structs.Allocation{a}))
require.Nil(state2.UpsertJob(999, a.Job))
require.Nil(state2.UpsertAllocs(1003, []*structs.Allocation{a}))
// Wait for the client to run the allocation
testutil.WaitForResult(func() (bool, error) {
alloc, err := state2.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
if alloc.ClientStatus != structs.AllocClientStatusComplete {
return false, fmt.Errorf("alloc client status: %v", alloc.ClientStatus)
}
return true, nil
}, func(err error) {
t.Fatalf("Alloc on node %q not finished: %v", c.NodeID(), err)
})
// Force remove the connection locally in case it exists
s1.nodeConnsLock.Lock()
delete(s1.nodeConns, c.NodeID())
s1.nodeConnsLock.Unlock()
// Make the request
req := &structs.AllocSpecificRequest{
AllocID: a.ID,
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp cstructs.ClientStatsResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.GarbageCollect", req, &resp)
require.Nil(err)
}
func TestClientAllocations_Stats_OldNode(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and fake an old client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
state := s.State()
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Test for an old version error
node := mock.Node()
node.Attributes["nomad.version"] = "0.7.1"
require.Nil(state.UpsertNode(1005, node))
alloc := mock.Alloc()
alloc.NodeID = node.ID
require.Nil(state.UpsertAllocs(1006, []*structs.Allocation{alloc}))
req := &structs.AllocSpecificRequest{
AllocID: alloc.ID,
QueryOptions: structs.QueryOptions{
Region: "global",
},
}
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.Stats", req, &resp)
require.True(structs.IsErrNodeLacksRpc(err), err.Error())
// Test for a missing version error
delete(node.Attributes, "nomad.version")
require.Nil(state.UpsertNode(1007, node))
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.Stats", req, &resp)
require.True(structs.IsErrUnknownNomadVersion(err), err.Error())
}
2018-02-06 01:20:42 +00:00
func TestClientAllocations_Stats_Local(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
2018-02-06 01:20:42 +00:00
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
c, cleanupC := client.TestClient(t, func(c *config.Config) {
2018-02-06 01:20:42 +00:00
c.Servers = []string{s.config.RPCAddr.String()}
})
defer cleanupC()
2018-02-06 01:20:42 +00:00
// Force an allocation onto the node
a := mock.Alloc()
a.Job.Type = structs.JobTypeBatch
a.NodeID = c.NodeID()
a.Job.TaskGroups[0].Count = 1
a.Job.TaskGroups[0].Tasks[0] = &structs.Task{
Name: "web",
Driver: "mock_driver",
Config: map[string]interface{}{
"run_for": "2s",
2018-02-06 01:20:42 +00:00
},
LogConfig: structs.DefaultLogConfig(),
Resources: &structs.Resources{
CPU: 500,
MemoryMB: 256,
},
}
testutil.WaitForResult(func() (bool, error) {
nodes := s.connectedNodes()
return len(nodes) == 1, nil
}, func(err error) {
t.Fatalf("should have a clients")
})
// Upsert the allocation
state := s.State()
require.Nil(state.UpsertJob(999, a.Job))
require.Nil(state.UpsertAllocs(1003, []*structs.Allocation{a}))
// Wait for the client to run the allocation
testutil.WaitForResult(func() (bool, error) {
alloc, err := state.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
if alloc.ClientStatus != structs.AllocClientStatusComplete {
return false, fmt.Errorf("alloc client status: %v", alloc.ClientStatus)
}
return true, nil
}, func(err error) {
t.Fatalf("Alloc on node %q not finished: %v", c.NodeID(), err)
})
// Make the request without having an alloc id
req := &structs.AllocSpecificRequest{
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp cstructs.AllocStatsResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.Stats", req, &resp)
require.NotNil(err)
require.EqualError(err, structs.ErrMissingAllocID.Error(), "(%T) %v")
2018-02-06 01:20:42 +00:00
// Fetch the response setting the node id
req.AllocID = a.ID
var resp2 cstructs.AllocStatsResponse
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.Stats", req, &resp2)
require.Nil(err)
require.NotNil(resp2.Stats)
}
func TestClientAllocations_Stats_Local_ACL(t *testing.T) {
t.Parallel()
// Start a server
s, root, cleanupS := TestACLServer(t, nil)
defer cleanupS()
2018-02-06 01:20:42 +00:00
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Create a bad token
policyBad := mock.NamespacePolicy("other", "", []string{acl.NamespaceCapabilityReadFS})
tokenBad := mock.CreatePolicyAndToken(t, s.State(), 1005, "invalid", policyBad)
policyGood := mock.NamespacePolicy(structs.DefaultNamespace, "", []string{acl.NamespaceCapabilityReadJob})
tokenGood := mock.CreatePolicyAndToken(t, s.State(), 1009, "valid2", policyGood)
// Upsert the allocation
state := s.State()
alloc := mock.Alloc()
require.NoError(t, state.UpsertJob(1010, alloc.Job))
require.NoError(t, state.UpsertAllocs(1011, []*structs.Allocation{alloc}))
2018-02-06 01:20:42 +00:00
cases := []struct {
Name string
Token string
ExpectedError string
}{
{
Name: "bad token",
Token: tokenBad.SecretID,
ExpectedError: structs.ErrPermissionDenied.Error(),
},
{
Name: "good token",
Token: tokenGood.SecretID,
ExpectedError: structs.ErrUnknownNodePrefix,
2018-02-06 01:20:42 +00:00
},
{
Name: "root token",
Token: root.SecretID,
ExpectedError: structs.ErrUnknownNodePrefix,
2018-02-06 01:20:42 +00:00
},
}
for _, c := range cases {
t.Run(c.Name, func(t *testing.T) {
// Make the request without having a node-id
req := &structs.AllocSpecificRequest{
AllocID: alloc.ID,
2018-02-06 01:20:42 +00:00
QueryOptions: structs.QueryOptions{
AuthToken: c.Token,
Region: "global",
Namespace: structs.DefaultNamespace,
},
}
// Fetch the response
var resp cstructs.AllocStatsResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.Stats", req, &resp)
require.NotNil(t, err)
require.Contains(t, err.Error(), c.ExpectedError)
2018-02-06 01:20:42 +00:00
})
}
}
func TestClientAllocations_Stats_Remote(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
s1, cleanupS1 := TestServer(t, func(c *Config) {
c.BootstrapExpect = 2
})
defer cleanupS1()
s2, cleanupS2 := TestServer(t, func(c *Config) {
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
c.BootstrapExpect = 2
2018-02-06 01:20:42 +00:00
})
defer cleanupS2()
2018-02-06 01:20:42 +00:00
TestJoin(t, s1, s2)
testutil.WaitForLeader(t, s1.RPC)
testutil.WaitForLeader(t, s2.RPC)
codec := rpcClient(t, s2)
c, cleanupC := client.TestClient(t, func(c *config.Config) {
2018-02-06 01:20:42 +00:00
c.Servers = []string{s2.config.RPCAddr.String()}
})
defer cleanupC()
2018-02-06 01:20:42 +00:00
// Force an allocation onto the node
a := mock.Alloc()
a.Job.Type = structs.JobTypeBatch
a.NodeID = c.NodeID()
a.Job.TaskGroups[0].Count = 1
a.Job.TaskGroups[0].Tasks[0] = &structs.Task{
Name: "web",
Driver: "mock_driver",
Config: map[string]interface{}{
"run_for": "2s",
2018-02-06 01:20:42 +00:00
},
LogConfig: structs.DefaultLogConfig(),
Resources: &structs.Resources{
CPU: 500,
MemoryMB: 256,
},
}
testutil.WaitForResult(func() (bool, error) {
nodes := s2.connectedNodes()
return len(nodes) == 1, nil
}, func(err error) {
t.Fatalf("should have a clients")
})
// Upsert the allocation
state1 := s1.State()
state2 := s2.State()
require.Nil(state1.UpsertJob(999, a.Job))
require.Nil(state1.UpsertAllocs(1003, []*structs.Allocation{a}))
require.Nil(state2.UpsertJob(999, a.Job))
require.Nil(state2.UpsertAllocs(1003, []*structs.Allocation{a}))
// Wait for the client to run the allocation
testutil.WaitForResult(func() (bool, error) {
alloc, err := state2.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
if alloc.ClientStatus != structs.AllocClientStatusComplete {
return false, fmt.Errorf("alloc client status: %v", alloc.ClientStatus)
}
return true, nil
}, func(err error) {
t.Fatalf("Alloc on node %q not finished: %v", c.NodeID(), err)
})
// Force remove the connection locally in case it exists
s1.nodeConnsLock.Lock()
delete(s1.nodeConns, c.NodeID())
s1.nodeConnsLock.Unlock()
// Make the request
req := &structs.AllocSpecificRequest{
AllocID: a.ID,
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp cstructs.AllocStatsResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.Stats", req, &resp)
require.Nil(err)
require.NotNil(resp.Stats)
}
func TestClientAllocations_Restart_Local(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
s, cleanupS := TestServer(t, nil)
defer cleanupS()
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
c, cleanupC := client.TestClient(t, func(c *config.Config) {
c.Servers = []string{s.config.RPCAddr.String()}
c.GCDiskUsageThreshold = 100.0
})
defer cleanupC()
// Force an allocation onto the node
a := mock.Alloc()
a.Job.Type = structs.JobTypeService
a.NodeID = c.NodeID()
a.Job.TaskGroups[0].Count = 1
a.Job.TaskGroups[0].Tasks[0] = &structs.Task{
Name: "web",
Driver: "mock_driver",
Config: map[string]interface{}{
"run_for": "10s",
},
LogConfig: structs.DefaultLogConfig(),
Resources: &structs.Resources{
CPU: 500,
MemoryMB: 256,
},
}
testutil.WaitForResult(func() (bool, error) {
nodes := s.connectedNodes()
return len(nodes) == 1, nil
}, func(err error) {
t.Fatalf("should have a client")
})
// Upsert the allocation
state := s.State()
require.Nil(state.UpsertJob(999, a.Job))
require.Nil(state.UpsertAllocs(1003, []*structs.Allocation{a}))
// Wait for the client to run the allocation
testutil.WaitForResult(func() (bool, error) {
alloc, err := state.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
if alloc.ClientStatus != structs.AllocClientStatusRunning {
return false, fmt.Errorf("alloc client status: %v", alloc.ClientStatus)
}
return true, nil
}, func(err error) {
t.Fatalf("Alloc on node %q not running: %v", c.NodeID(), err)
})
// Make the request without having an alloc id
req := &structs.AllocRestartRequest{
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.Restart", req, &resp)
require.NotNil(err)
require.EqualError(err, structs.ErrMissingAllocID.Error(), "(%T) %v")
// Fetch the response setting the alloc id - This should not error because the
// alloc is running.
req.AllocID = a.ID
var resp2 structs.GenericResponse
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.Restart", req, &resp2)
require.Nil(err)
testutil.WaitForResult(func() (bool, error) {
alloc, err := state.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
taskState := alloc.TaskStates["web"]
if taskState == nil {
return false, fmt.Errorf("could not find task state")
}
if taskState.Restarts != 1 {
return false, fmt.Errorf("expected task 'web' to have 1 restart, got: %d", taskState.Restarts)
}
return true, nil
}, func(err error) {
t.Fatalf("Alloc on node %q not running: %v", c.NodeID(), err)
})
}
func TestClientAllocations_Restart_Remote(t *testing.T) {
t.Parallel()
require := require.New(t)
// Start a server and client
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
s1, cleanupS1 := TestServer(t, func(c *Config) {
c.BootstrapExpect = 2
})
defer cleanupS1()
s2, cleanupS2 := TestServer(t, func(c *Config) {
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
c.BootstrapExpect = 2
})
defer cleanupS2()
TestJoin(t, s1, s2)
testutil.WaitForLeader(t, s1.RPC)
testutil.WaitForLeader(t, s2.RPC)
codec := rpcClient(t, s2)
c, cleanupC := client.TestClient(t, func(c *config.Config) {
c.Servers = []string{s2.config.RPCAddr.String()}
})
defer cleanupC()
// Force an allocation onto the node
a := mock.Alloc()
a.Job.Type = structs.JobTypeService
a.NodeID = c.NodeID()
a.Job.TaskGroups[0].Count = 1
a.Job.TaskGroups[0].Tasks[0] = &structs.Task{
Name: "web",
Driver: "mock_driver",
Config: map[string]interface{}{
"run_for": "10s",
},
LogConfig: structs.DefaultLogConfig(),
Resources: &structs.Resources{
CPU: 500,
MemoryMB: 256,
},
}
testutil.WaitForResult(func() (bool, error) {
nodes := s2.connectedNodes()
return len(nodes) == 1, nil
}, func(err error) {
t.Fatalf("should have a client")
})
// Upsert the allocation
state1 := s1.State()
state2 := s2.State()
require.Nil(state1.UpsertJob(999, a.Job))
require.Nil(state1.UpsertAllocs(1003, []*structs.Allocation{a}))
require.Nil(state2.UpsertJob(999, a.Job))
require.Nil(state2.UpsertAllocs(1003, []*structs.Allocation{a}))
// Wait for the client to run the allocation
testutil.WaitForResult(func() (bool, error) {
alloc, err := state2.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
if alloc.ClientStatus != structs.AllocClientStatusRunning {
return false, fmt.Errorf("alloc client status: %v", alloc.ClientStatus)
}
return true, nil
}, func(err error) {
t.Fatalf("Alloc on node %q not running: %v", c.NodeID(), err)
})
// Make the request without having an alloc id
req := &structs.AllocRestartRequest{
QueryOptions: structs.QueryOptions{Region: "global"},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.Restart", req, &resp)
require.NotNil(err)
require.EqualError(err, structs.ErrMissingAllocID.Error(), "(%T) %v")
// Fetch the response setting the alloc id - This should succeed because the
// alloc is running
req.AllocID = a.ID
var resp2 structs.GenericResponse
err = msgpackrpc.CallWithCodec(codec, "ClientAllocations.Restart", req, &resp2)
require.NoError(err)
}
func TestClientAllocations_Restart_ACL(t *testing.T) {
// Start a server
s, root, cleanupS := TestACLServer(t, nil)
defer cleanupS()
codec := rpcClient(t, s)
testutil.WaitForLeader(t, s.RPC)
// Create a bad token
policyBad := mock.NamespacePolicy("other", "", []string{acl.NamespaceCapabilityReadFS})
tokenBad := mock.CreatePolicyAndToken(t, s.State(), 1005, "invalid", policyBad)
policyGood := mock.NamespacePolicy(structs.DefaultNamespace, acl.PolicyWrite, nil)
tokenGood := mock.CreatePolicyAndToken(t, s.State(), 1009, "valid2", policyGood)
// Upsert the allocation
state := s.State()
alloc := mock.Alloc()
require.NoError(t, state.UpsertJob(1010, alloc.Job))
require.NoError(t, state.UpsertAllocs(1011, []*structs.Allocation{alloc}))
cases := []struct {
Name string
Token string
ExpectedError string
}{
{
Name: "bad token",
Token: tokenBad.SecretID,
ExpectedError: structs.ErrPermissionDenied.Error(),
},
{
Name: "good token",
Token: tokenGood.SecretID,
ExpectedError: "Unknown node",
},
{
Name: "root token",
Token: root.SecretID,
ExpectedError: "Unknown node",
},
}
for _, c := range cases {
t.Run(c.Name, func(t *testing.T) {
// Make the request without having a node-id
req := &structs.AllocRestartRequest{
AllocID: alloc.ID,
QueryOptions: structs.QueryOptions{
Namespace: structs.DefaultNamespace,
AuthToken: c.Token,
Region: "global",
},
}
// Fetch the response
var resp structs.GenericResponse
err := msgpackrpc.CallWithCodec(codec, "ClientAllocations.Restart", req, &resp)
require.NotNil(t, err)
require.Contains(t, err.Error(), c.ExpectedError)
})
}
}
// TestAlloc_ExecStreaming asserts that exec task requests are forwarded
// to appropriate server or remote regions
func TestAlloc_ExecStreaming(t *testing.T) {
t.Parallel()
////// Nomad clusters topology - not specific to test
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
localServer, cleanupLS := TestServer(t, func(c *Config) {
c.BootstrapExpect = 2
})
defer cleanupLS()
remoteServer, cleanupRS := TestServer(t, func(c *Config) {
Simplify Bootstrap logic in tests This change updates tests to honor `BootstrapExpect` exclusively when forming test clusters and removes test only knobs, e.g. `config.DevDisableBootstrap`. Background: Test cluster creation is fragile. Test servers don't follow the BootstapExpected route like production clusters. Instead they start as single node clusters and then get rejoin and may risk causing brain split or other test flakiness. The test framework expose few knobs to control those (e.g. `config.DevDisableBootstrap` and `config.Bootstrap`) that control whether a server should bootstrap the cluster. These flags are confusing and it's unclear when to use: their usage in multi-node cluster isn't properly documented. Furthermore, they have some bad side-effects as they don't control Raft library: If `config.DevDisableBootstrap` is true, the test server may not immediately attempt to bootstrap a cluster, but after an election timeout (~50ms), Raft may force a leadership election and win it (with only one vote) and cause a split brain. The knobs are also confusing as Bootstrap is an overloaded term. In BootstrapExpect, we refer to bootstrapping the cluster only after N servers are connected. But in tests and the knobs above, it refers to whether the server is a single node cluster and shouldn't wait for any other server. Changes: This commit makes two changes: First, it relies on `BootstrapExpected` instead of `Bootstrap` and/or `DevMode` flags. This change is relatively trivial. Introduce a `Bootstrapped` flag to track if the cluster is bootstrapped. This allows us to keep `BootstrapExpected` immutable. Previously, the flag was a config value but it gets set to 0 after cluster bootstrap completes.
2020-03-02 15:29:24 +00:00
c.BootstrapExpect = 2
})
defer cleanupRS()
remoteRegionServer, cleanupRRS := TestServer(t, func(c *Config) {
c.Region = "two"
})
defer cleanupRRS()
TestJoin(t, localServer, remoteServer)
TestJoin(t, localServer, remoteRegionServer)
testutil.WaitForLeader(t, localServer.RPC)
testutil.WaitForLeader(t, remoteServer.RPC)
testutil.WaitForLeader(t, remoteRegionServer.RPC)
c, cleanup := client.TestClient(t, func(c *config.Config) {
c.Servers = []string{localServer.config.RPCAddr.String()}
})
defer cleanup()
// Wait for the client to connect
testutil.WaitForResult(func() (bool, error) {
nodes := remoteServer.connectedNodes()
return len(nodes) == 1, nil
}, func(err error) {
require.NoError(t, err, "failed to have a client")
})
// Force remove the connection locally in case it exists
remoteServer.nodeConnsLock.Lock()
delete(remoteServer.nodeConns, c.NodeID())
remoteServer.nodeConnsLock.Unlock()
///// Start task
a := mock.BatchAlloc()
a.NodeID = c.NodeID()
a.Job.Type = structs.JobTypeBatch
a.Job.TaskGroups[0].Count = 1
a.Job.TaskGroups[0].Tasks[0].Config = map[string]interface{}{
"run_for": "20s",
"exec_command": map[string]interface{}{
"run_for": "1ms",
"stdout_string": "expected output",
"exit_code": 3,
},
}
// Upsert the allocation
localState := localServer.State()
require.Nil(t, localState.UpsertJob(999, a.Job))
require.Nil(t, localState.UpsertAllocs(1003, []*structs.Allocation{a}))
remoteState := remoteServer.State()
require.Nil(t, remoteState.UpsertJob(999, a.Job))
require.Nil(t, remoteState.UpsertAllocs(1003, []*structs.Allocation{a}))
// Wait for the client to run the allocation
testutil.WaitForResult(func() (bool, error) {
alloc, err := localState.AllocByID(nil, a.ID)
if err != nil {
return false, err
}
if alloc == nil {
return false, fmt.Errorf("unknown alloc")
}
if alloc.ClientStatus != structs.AllocClientStatusRunning {
return false, fmt.Errorf("alloc client status: %v", alloc.ClientStatus)
}
return true, nil
}, func(err error) {
require.NoError(t, err, "task didn't start yet")
})
///////// Actually run query now
cases := []struct {
name string
rpc func(string) (structs.StreamingRpcHandler, error)
}{
{"client", c.StreamingRpcHandler},
{"local_server", localServer.StreamingRpcHandler},
{"remote_server", remoteServer.StreamingRpcHandler},
{"remote_region", remoteRegionServer.StreamingRpcHandler},
}
for _, tc := range cases {
t.Run(tc.name, func(t *testing.T) {
// Make the request
req := &cstructs.AllocExecRequest{
AllocID: a.ID,
Task: a.Job.TaskGroups[0].Tasks[0].Name,
Tty: true,
Cmd: []string{"placeholder command"},
QueryOptions: nstructs.QueryOptions{Region: "global"},
}
// Get the handler
handler, err := tc.rpc("Allocations.Exec")
require.Nil(t, err)
// Create a pipe
p1, p2 := net.Pipe()
defer p1.Close()
defer p2.Close()
errCh := make(chan error)
frames := make(chan *drivers.ExecTaskStreamingResponseMsg)
// Start the handler
go handler(p2)
go decodeFrames(t, p1, frames, errCh)
// Send the request
encoder := codec.NewEncoder(p1, nstructs.MsgpackHandle)
require.Nil(t, encoder.Encode(req))
timeout := time.After(3 * time.Second)
OUTER:
for {
select {
case <-timeout:
require.FailNow(t, "timed out before getting exit code")
case err := <-errCh:
require.NoError(t, err)
case f := <-frames:
if f.Exited && f.Result != nil {
code := int(f.Result.ExitCode)
require.Equal(t, 3, code)
break OUTER
}
}
}
})
}
}
func decodeFrames(t *testing.T, p1 net.Conn, frames chan<- *drivers.ExecTaskStreamingResponseMsg, errCh chan<- error) {
// Start the decoder
decoder := codec.NewDecoder(p1, nstructs.MsgpackHandle)
for {
var msg cstructs.StreamErrWrapper
if err := decoder.Decode(&msg); err != nil {
if err == io.EOF || strings.Contains(err.Error(), "closed") {
return
}
t.Logf("received error decoding: %#v", err)
errCh <- fmt.Errorf("error decoding: %v", err)
return
}
if msg.Error != nil {
errCh <- msg.Error
continue
}
var frame drivers.ExecTaskStreamingResponseMsg
json.Unmarshal(msg.Payload, &frame)
t.Logf("received message: %#v", msg)
frames <- &frame
}
}