open-nomad/nomad/heartbeat_test.go

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

360 lines
8.8 KiB
Go
Raw Normal View History

// Copyright (c) HashiCorp, Inc.
// SPDX-License-Identifier: MPL-2.0
2015-08-23 00:37:50 +00:00
package nomad
import (
"fmt"
"testing"
"time"
2017-02-08 05:22:48 +00:00
memdb "github.com/hashicorp/go-memdb"
msgpackrpc "github.com/hashicorp/net-rpc-msgpackrpc"
"github.com/hashicorp/nomad/ci"
"github.com/hashicorp/nomad/helper/pointer"
2015-08-23 00:37:50 +00:00
"github.com/hashicorp/nomad/nomad/mock"
"github.com/hashicorp/nomad/nomad/structs"
"github.com/hashicorp/nomad/testutil"
2018-02-20 18:22:15 +00:00
"github.com/stretchr/testify/require"
2015-08-23 00:37:50 +00:00
)
2018-02-20 18:22:15 +00:00
func TestHeartbeat_InitializeHeartbeatTimers(t *testing.T) {
ci.Parallel(t)
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
2015-08-23 00:37:50 +00:00
testutil.WaitForLeader(t, s1.RPC)
node := mock.Node()
state := s1.fsm.State()
err := state.UpsertNode(structs.MsgTypeTestSetup, 1, node)
2015-08-23 00:37:50 +00:00
if err != nil {
t.Fatalf("err: %v", err)
}
// Reset the heartbeat timers
err = s1.initializeHeartbeatTimers()
if err != nil {
t.Fatalf("err: %v", err)
}
// Check that we have a timer
_, ok := s1.heartbeatTimers[node.ID]
if !ok {
t.Fatalf("missing heartbeat timer")
}
}
2018-02-20 18:22:15 +00:00
func TestHeartbeat_ResetHeartbeatTimer(t *testing.T) {
ci.Parallel(t)
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
2015-08-23 00:37:50 +00:00
testutil.WaitForLeader(t, s1.RPC)
// Create a new timer
ttl, err := s1.resetHeartbeatTimer("test")
if err != nil {
t.Fatalf("err: %v", err)
}
2015-08-31 01:10:12 +00:00
if ttl < s1.config.MinHeartbeatTTL || ttl > 2*s1.config.MinHeartbeatTTL {
2015-08-23 00:37:50 +00:00
t.Fatalf("bad: %#v", ttl)
}
// Check that we have a timer
_, ok := s1.heartbeatTimers["test"]
if !ok {
t.Fatalf("missing heartbeat timer")
}
}
2018-02-20 18:22:15 +00:00
func TestHeartbeat_ResetHeartbeatTimer_Nonleader(t *testing.T) {
ci.Parallel(t)
2018-02-20 18:22:15 +00:00
require := require.New(t)
s1, cleanupS1 := TestServer(t, func(c *Config) {
2018-02-20 18:22:15 +00:00
c.BootstrapExpect = 3 // Won't become leader
})
defer cleanupS1()
2018-02-20 18:22:15 +00:00
require.False(s1.IsLeader())
// Create a new timer
_, err := s1.resetHeartbeatTimer("test")
require.NotNil(err)
require.EqualError(err, heartbeatNotLeader)
}
func TestHeartbeat_ResetHeartbeatTimerLocked(t *testing.T) {
ci.Parallel(t)
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
2015-08-23 00:37:50 +00:00
testutil.WaitForLeader(t, s1.RPC)
s1.heartbeatTimersLock.Lock()
s1.resetHeartbeatTimerLocked("foo", 5*time.Millisecond)
s1.heartbeatTimersLock.Unlock()
if _, ok := s1.heartbeatTimers["foo"]; !ok {
t.Fatalf("missing timer")
}
2017-01-26 23:06:14 +00:00
time.Sleep(time.Duration(testutil.TestMultiplier()*10) * time.Millisecond)
2015-08-23 00:37:50 +00:00
if _, ok := s1.heartbeatTimers["foo"]; ok {
t.Fatalf("timer should be gone")
}
}
2018-02-20 18:22:15 +00:00
func TestHeartbeat_ResetHeartbeatTimerLocked_Renew(t *testing.T) {
ci.Parallel(t)
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
2015-08-23 00:37:50 +00:00
testutil.WaitForLeader(t, s1.RPC)
s1.heartbeatTimersLock.Lock()
2017-10-20 00:13:27 +00:00
s1.resetHeartbeatTimerLocked("foo", 30*time.Millisecond)
2015-08-23 00:37:50 +00:00
s1.heartbeatTimersLock.Unlock()
if _, ok := s1.heartbeatTimers["foo"]; !ok {
t.Fatalf("missing timer")
}
time.Sleep(2 * time.Millisecond)
// Renew the heartbeat
s1.heartbeatTimersLock.Lock()
2017-10-20 00:13:27 +00:00
s1.resetHeartbeatTimerLocked("foo", 30*time.Millisecond)
2015-08-23 00:37:50 +00:00
s1.heartbeatTimersLock.Unlock()
renew := time.Now()
// Watch for invalidation
2017-10-20 00:13:27 +00:00
for time.Now().Sub(renew) < time.Duration(testutil.TestMultiplier()*100)*time.Millisecond {
2015-08-23 00:37:50 +00:00
s1.heartbeatTimersLock.Lock()
_, ok := s1.heartbeatTimers["foo"]
s1.heartbeatTimersLock.Unlock()
if !ok {
end := time.Now()
2017-10-20 00:13:27 +00:00
if diff := end.Sub(renew); diff < 30*time.Millisecond {
2015-08-23 00:37:50 +00:00
t.Fatalf("early invalidate %v", diff)
}
return
}
2017-10-20 00:13:27 +00:00
time.Sleep(2 * time.Millisecond)
2015-08-23 00:37:50 +00:00
}
t.Fatalf("should have expired")
}
2018-02-20 18:22:15 +00:00
func TestHeartbeat_InvalidateHeartbeat(t *testing.T) {
ci.Parallel(t)
2018-05-11 21:53:41 +00:00
require := require.New(t)
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
2015-08-23 00:37:50 +00:00
testutil.WaitForLeader(t, s1.RPC)
// Create a node
node := mock.Node()
state := s1.fsm.State()
require.NoError(state.UpsertNode(structs.MsgTypeTestSetup, 1, node))
2015-08-23 00:37:50 +00:00
// This should cause a status update
s1.invalidateHeartbeat(node.ID)
// Check it is updated
2017-02-08 05:22:48 +00:00
ws := memdb.NewWatchSet()
out, err := state.NodeByID(ws, node.ID)
2018-05-11 21:53:41 +00:00
require.NoError(err)
require.True(out.TerminalStatus())
require.Len(out.Events, 2)
require.Equal(NodeHeartbeatEventMissed, out.Events[1].Message)
2015-08-23 00:37:50 +00:00
}
2018-02-20 18:22:15 +00:00
func TestHeartbeat_ClearHeartbeatTimer(t *testing.T) {
ci.Parallel(t)
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
2015-08-23 00:37:50 +00:00
testutil.WaitForLeader(t, s1.RPC)
s1.heartbeatTimersLock.Lock()
s1.resetHeartbeatTimerLocked("foo", 5*time.Millisecond)
s1.heartbeatTimersLock.Unlock()
err := s1.clearHeartbeatTimer("foo")
if err != nil {
t.Fatalf("err: %v", err)
}
if _, ok := s1.heartbeatTimers["foo"]; ok {
t.Fatalf("timer should be gone")
}
}
2018-02-20 18:22:15 +00:00
func TestHeartbeat_ClearAllHeartbeatTimers(t *testing.T) {
ci.Parallel(t)
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
2015-08-23 00:37:50 +00:00
testutil.WaitForLeader(t, s1.RPC)
s1.heartbeatTimersLock.Lock()
s1.resetHeartbeatTimerLocked("foo", 10*time.Millisecond)
s1.resetHeartbeatTimerLocked("bar", 10*time.Millisecond)
s1.resetHeartbeatTimerLocked("baz", 10*time.Millisecond)
s1.heartbeatTimersLock.Unlock()
err := s1.clearAllHeartbeatTimers()
if err != nil {
t.Fatalf("err: %v", err)
}
if len(s1.heartbeatTimers) != 0 {
t.Fatalf("timers should be gone")
}
}
2018-02-20 18:22:15 +00:00
func TestHeartbeat_Server_HeartbeatTTL_Failover(t *testing.T) {
ci.Parallel(t)
2015-08-23 00:37:50 +00:00
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 = 3
})
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 = 3
2015-08-23 00:37:50 +00:00
})
defer cleanupS2()
2015-08-23 00:37:50 +00:00
s3, cleanupS3 := 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 = 3
2015-08-23 00:37:50 +00:00
})
defer cleanupS3()
2015-08-23 00:37:50 +00:00
servers := []*Server{s1, s2, s3}
2018-01-12 01:00:30 +00:00
TestJoin(t, s1, s2, s3)
2015-08-23 00:37:50 +00:00
leader := waitForStableLeadership(t, servers)
2015-08-23 00:37:50 +00:00
codec := rpcClient(t, leader)
// Create the register request
node := mock.Node()
req := &structs.NodeRegisterRequest{
Node: node,
WriteRequest: structs.WriteRequest{Region: "global"},
2015-08-23 00:37:50 +00:00
}
// Fetch the response
var resp structs.GenericResponse
if err := msgpackrpc.CallWithCodec(codec, "Node.Register", req, &resp); err != nil {
2015-08-23 00:37:50 +00:00
t.Fatalf("err: %v", err)
}
// Check that heartbeatTimers has the heartbeat ID
if _, ok := leader.heartbeatTimers[node.ID]; !ok {
t.Fatalf("missing heartbeat timer")
}
// Shutdown the leader!
leader.Shutdown()
// heartbeatTimers should be cleared on leader shutdown
2018-02-21 00:05:43 +00:00
testutil.WaitForResult(func() (bool, error) {
return len(leader.heartbeatTimers) == 0, nil
}, func(err error) {
2015-08-23 00:37:50 +00:00
t.Fatalf("heartbeat timers should be empty on the shutdown leader")
2018-02-21 00:05:43 +00:00
})
2015-08-23 00:37:50 +00:00
// Find the new leader
testutil.WaitForResult(func() (bool, error) {
leader = nil
for _, s := range servers {
if s.IsLeader() {
leader = s
}
}
if leader == nil {
return false, fmt.Errorf("Should have a new leader")
}
// Ensure heartbeat timer is restored
if _, ok := leader.heartbeatTimers[node.ID]; !ok {
return false, fmt.Errorf("missing heartbeat timer")
}
return true, nil
}, func(err error) {
t.Fatalf("err: %s", err)
})
}
func TestHeartbeat_InvalidateHeartbeat_DisconnectedClient(t *testing.T) {
ci.Parallel(t)
type testCase struct {
name string
now time.Time
maxClientDisconnect *time.Duration
expectedNodeStatus string
}
testCases := []testCase{
{
name: "has-pending-reconnects",
now: time.Now().UTC(),
maxClientDisconnect: pointer.Of(5 * time.Second),
expectedNodeStatus: structs.NodeStatusDisconnected,
},
{
name: "has-expired-reconnects",
maxClientDisconnect: pointer.Of(5 * time.Second),
now: time.Now().UTC().Add(-10 * time.Second),
expectedNodeStatus: structs.NodeStatusDown,
},
{
name: "has-expired-reconnects-equal-timestamp",
maxClientDisconnect: pointer.Of(5 * time.Second),
now: time.Now().UTC().Add(-5 * time.Second),
expectedNodeStatus: structs.NodeStatusDown,
},
{
name: "has-no-reconnects",
now: time.Now().UTC(),
maxClientDisconnect: nil,
expectedNodeStatus: structs.NodeStatusDown,
},
}
for _, tc := range testCases {
t.Run(tc.name, func(t *testing.T) {
s1, cleanupS1 := TestServer(t, nil)
defer cleanupS1()
testutil.WaitForLeader(t, s1.RPC)
// Create a node
node := mock.Node()
state := s1.fsm.State()
require.NoError(t, state.UpsertNode(structs.MsgTypeTestSetup, 1, node))
alloc := mock.Alloc()
alloc.NodeID = node.ID
alloc.Job.TaskGroups[0].MaxClientDisconnect = tc.maxClientDisconnect
alloc.ClientStatus = structs.AllocClientStatusUnknown
alloc.AllocStates = []*structs.AllocState{{
Field: structs.AllocStateFieldClientStatus,
Value: structs.AllocClientStatusUnknown,
Time: tc.now,
}}
require.NoError(t, state.UpsertAllocs(structs.MsgTypeTestSetup, 2, []*structs.Allocation{alloc}))
// Trigger status update
s1.invalidateHeartbeat(node.ID)
out, err := state.NodeByID(nil, node.ID)
require.NoError(t, err)
require.Equal(t, tc.expectedNodeStatus, out.Status)
})
}
}