mirror of
https://github.com/kemko/nomad.git
synced 2026-01-01 16:05:42 +03:00
This PR introduces support for using Nomad on systems with cgroups v2 [1] enabled as the cgroups controller mounted on /sys/fs/cgroups. Newer Linux distros like Ubuntu 21.10 are shipping with cgroups v2 only, causing problems for Nomad users. Nomad mostly "just works" with cgroups v2 due to the indirection via libcontainer, but not so for managing cpuset cgroups. Before, Nomad has been making use of a feature in v1 where a PID could be a member of more than one cgroup. In v2 this is no longer possible, and so the logic around computing cpuset values must be modified. When Nomad detects v2, it manages cpuset values in-process, rather than making use of cgroup heirarchy inheritence via shared/reserved parents. Nomad will only activate the v2 logic when it detects cgroups2 is mounted at /sys/fs/cgroups. This means on systems running in hybrid mode with cgroups2 mounted at /sys/fs/cgroups/unified (as is typical) Nomad will continue to use the v1 logic, and should operate as before. Systems that do not support cgroups v2 are also not affected. When v2 is activated, Nomad will create a parent called nomad.slice (unless otherwise configured in Client conifg), and create cgroups for tasks using naming convention <allocID>-<task>.scope. These follow the naming convention set by systemd and also used by Docker when cgroups v2 is detected. Client nodes now export a new fingerprint attribute, unique.cgroups.version which will be set to 'v1' or 'v2' to indicate the cgroups regime in use by Nomad. The new cpuset management strategy fixes #11705, where docker tasks that spawned processes on startup would "leak". In cgroups v2, the PIDs are started in the cgroup they will always live in, and thus the cause of the leak is eliminated. [1] https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html Closes #11289 Fixes #11705 #11773 #11933
348 lines
11 KiB
Go
348 lines
11 KiB
Go
package testutils
|
|
|
|
import (
|
|
"context"
|
|
"fmt"
|
|
"io/ioutil"
|
|
"os"
|
|
"path/filepath"
|
|
"runtime"
|
|
"strings"
|
|
"time"
|
|
|
|
hclog "github.com/hashicorp/go-hclog"
|
|
plugin "github.com/hashicorp/go-plugin"
|
|
"github.com/hashicorp/nomad/client/allocdir"
|
|
"github.com/hashicorp/nomad/client/config"
|
|
"github.com/hashicorp/nomad/client/lib/cgutil"
|
|
"github.com/hashicorp/nomad/client/logmon"
|
|
"github.com/hashicorp/nomad/client/taskenv"
|
|
"github.com/hashicorp/nomad/helper/testlog"
|
|
"github.com/hashicorp/nomad/helper/uuid"
|
|
"github.com/hashicorp/nomad/nomad/mock"
|
|
"github.com/hashicorp/nomad/nomad/structs"
|
|
"github.com/hashicorp/nomad/plugins/base"
|
|
"github.com/hashicorp/nomad/plugins/drivers"
|
|
"github.com/hashicorp/nomad/plugins/shared/hclspec"
|
|
testing "github.com/mitchellh/go-testing-interface"
|
|
"github.com/stretchr/testify/require"
|
|
)
|
|
|
|
type DriverHarness struct {
|
|
drivers.DriverPlugin
|
|
client *plugin.GRPCClient
|
|
server *plugin.GRPCServer
|
|
t testing.T
|
|
logger hclog.Logger
|
|
impl drivers.DriverPlugin
|
|
cgroup string
|
|
}
|
|
|
|
func (h *DriverHarness) Impl() drivers.DriverPlugin {
|
|
return h.impl
|
|
}
|
|
func NewDriverHarness(t testing.T, d drivers.DriverPlugin) *DriverHarness {
|
|
logger := testlog.HCLogger(t).Named("driver_harness")
|
|
|
|
pd := drivers.NewDriverPlugin(d, logger)
|
|
|
|
client, server := plugin.TestPluginGRPCConn(t,
|
|
map[string]plugin.Plugin{
|
|
base.PluginTypeDriver: pd,
|
|
base.PluginTypeBase: &base.PluginBase{Impl: d},
|
|
"logmon": logmon.NewPlugin(logmon.NewLogMon(logger.Named("logmon"))),
|
|
},
|
|
)
|
|
|
|
raw, err := client.Dispense(base.PluginTypeDriver)
|
|
require.NoError(t, err, "failed to dispense plugin")
|
|
|
|
dClient := raw.(drivers.DriverPlugin)
|
|
return &DriverHarness{
|
|
client: client,
|
|
server: server,
|
|
DriverPlugin: dClient,
|
|
logger: logger,
|
|
t: t,
|
|
impl: d,
|
|
}
|
|
}
|
|
|
|
// setCgroup creates a v2 cgroup for the task, as if a Client were initialized
|
|
// and managing the cgroup as it normally would.
|
|
//
|
|
// Uses testing.slice as a parent.
|
|
func (h *DriverHarness) setCgroup(allocID, task string) {
|
|
if cgutil.UseV2 {
|
|
h.cgroup = filepath.Join(cgutil.CgroupRoot, "testing.slice", cgutil.CgroupScope(allocID, task))
|
|
f, err := os.Create(h.cgroup)
|
|
if err != nil {
|
|
panic(err)
|
|
}
|
|
defer f.Close()
|
|
}
|
|
}
|
|
|
|
func (h *DriverHarness) Kill() {
|
|
_ = h.client.Close()
|
|
h.server.Stop()
|
|
h.cleanupCgroup()
|
|
}
|
|
|
|
func (h *DriverHarness) cleanupCgroup() {
|
|
if cgutil.UseV2 {
|
|
err := os.Remove(h.cgroup)
|
|
if err != nil {
|
|
panic(err)
|
|
}
|
|
}
|
|
}
|
|
|
|
// tinyChroot is useful for testing, where we do not use anything other than
|
|
// trivial /bin commands like sleep and sh.
|
|
//
|
|
// Note that you cannot chroot a symlink.
|
|
var tinyChroot = map[string]string{
|
|
// destination: /bin
|
|
"/usr/bin/sleep": "/bin/sleep",
|
|
"/usr/bin/dash": "/bin/sh",
|
|
"/usr/bin/bash": "/bin/bash",
|
|
"/usr/bin/cat": "/bin/cat",
|
|
|
|
// destination: /usr/bin
|
|
"/usr/bin/stty": "/usr/bin/stty",
|
|
"/usr/bin/head": "/usr/bin/head",
|
|
"/usr/bin/mktemp": "/usr/bin/mktemp",
|
|
"/usr/bin/echo": "/usr/bin/echo",
|
|
"/usr/bin/touch": "/usr/bin/touch",
|
|
"/usr/bin/stat": "/usr/bin/stat",
|
|
|
|
// destination: /etc/
|
|
"/etc/ld.so.cache": "/etc/ld.so.cache",
|
|
"/etc/ld.so.conf": "/etc/ld.so.conf",
|
|
"/etc/ld.so.conf.d": "/etc/ld.so.conf.d",
|
|
"/etc/passwd": "/etc/passwd",
|
|
"/etc/resolv.conf": "/etc/resolv.conf",
|
|
|
|
// others
|
|
"/lib": "/lib",
|
|
"/lib32": "/lib32",
|
|
"/lib64": "/lib64",
|
|
"/usr/lib/jvm": "/usr/lib/jvm",
|
|
"/run/resolvconf": "/run/resolvconf",
|
|
"/run/systemd/resolve": "/run/systemd/resolve",
|
|
}
|
|
|
|
// MkAllocDir creates a temporary directory and allocdir structure.
|
|
// If enableLogs is set to true a logmon instance will be started to write logs
|
|
// to the LogDir of the task
|
|
// A cleanup func is returned and should be deferred so as to not leak dirs
|
|
// between tests.
|
|
func (h *DriverHarness) MkAllocDir(t *drivers.TaskConfig, enableLogs bool) func() {
|
|
dir, err := ioutil.TempDir("", "nomad_driver_harness-")
|
|
require.NoError(h.t, err)
|
|
|
|
allocDir := allocdir.NewAllocDir(h.logger, dir, t.AllocID)
|
|
require.NoError(h.t, allocDir.Build())
|
|
|
|
t.AllocDir = allocDir.AllocDir
|
|
|
|
taskDir := allocDir.NewTaskDir(t.Name)
|
|
|
|
caps, err := h.Capabilities()
|
|
require.NoError(h.t, err)
|
|
|
|
fsi := caps.FSIsolation
|
|
require.NoError(h.t, taskDir.Build(fsi == drivers.FSIsolationChroot, tinyChroot))
|
|
|
|
task := &structs.Task{
|
|
Name: t.Name,
|
|
Env: t.Env,
|
|
}
|
|
|
|
// Create the mock allocation
|
|
alloc := mock.Alloc()
|
|
alloc.ID = t.AllocID
|
|
if t.Resources != nil {
|
|
alloc.AllocatedResources.Tasks[task.Name] = t.Resources.NomadResources
|
|
}
|
|
|
|
taskBuilder := taskenv.NewBuilder(mock.Node(), alloc, task, "global")
|
|
SetEnvvars(taskBuilder, fsi, taskDir, config.DefaultConfig())
|
|
|
|
taskEnv := taskBuilder.Build()
|
|
if t.Env == nil {
|
|
t.Env = taskEnv.Map()
|
|
} else {
|
|
for k, v := range taskEnv.Map() {
|
|
if _, ok := t.Env[k]; !ok {
|
|
t.Env[k] = v
|
|
}
|
|
}
|
|
}
|
|
|
|
// set cgroup
|
|
h.setCgroup(alloc.ID, task.Name)
|
|
|
|
//logmon
|
|
if enableLogs {
|
|
lm := logmon.NewLogMon(h.logger.Named("logmon"))
|
|
if runtime.GOOS == "windows" {
|
|
id := uuid.Generate()[:8]
|
|
t.StdoutPath = fmt.Sprintf("//./pipe/%s-%s.stdout", t.Name, id)
|
|
t.StderrPath = fmt.Sprintf("//./pipe/%s-%s.stderr", t.Name, id)
|
|
} else {
|
|
t.StdoutPath = filepath.Join(taskDir.LogDir, fmt.Sprintf(".%s.stdout.fifo", t.Name))
|
|
t.StderrPath = filepath.Join(taskDir.LogDir, fmt.Sprintf(".%s.stderr.fifo", t.Name))
|
|
}
|
|
err = lm.Start(&logmon.LogConfig{
|
|
LogDir: taskDir.LogDir,
|
|
StdoutLogFile: fmt.Sprintf("%s.stdout", t.Name),
|
|
StderrLogFile: fmt.Sprintf("%s.stderr", t.Name),
|
|
StdoutFifo: t.StdoutPath,
|
|
StderrFifo: t.StderrPath,
|
|
MaxFiles: 10,
|
|
MaxFileSizeMB: 10,
|
|
})
|
|
require.NoError(h.t, err)
|
|
|
|
return func() {
|
|
lm.Stop()
|
|
h.client.Close()
|
|
allocDir.Destroy()
|
|
}
|
|
}
|
|
|
|
return func() {
|
|
h.client.Close()
|
|
allocDir.Destroy()
|
|
h.cleanupCgroup()
|
|
}
|
|
}
|
|
|
|
// WaitUntilStarted will block until the task for the given ID is in the running
|
|
// state or the timeout is reached
|
|
func (h *DriverHarness) WaitUntilStarted(taskID string, timeout time.Duration) error {
|
|
deadline := time.Now().Add(timeout)
|
|
var lastState drivers.TaskState
|
|
for {
|
|
status, err := h.InspectTask(taskID)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
if status.State == drivers.TaskStateRunning {
|
|
return nil
|
|
}
|
|
lastState = status.State
|
|
if time.Now().After(deadline) {
|
|
return fmt.Errorf("task never transitioned to running, currently '%s'", lastState)
|
|
}
|
|
time.Sleep(100 * time.Millisecond)
|
|
}
|
|
}
|
|
|
|
// MockDriver is used for testing.
|
|
// Each function can be set as a closure to make assertions about how data
|
|
// is passed through the base plugin layer.
|
|
type MockDriver struct {
|
|
base.MockPlugin
|
|
TaskConfigSchemaF func() (*hclspec.Spec, error)
|
|
FingerprintF func(context.Context) (<-chan *drivers.Fingerprint, error)
|
|
CapabilitiesF func() (*drivers.Capabilities, error)
|
|
RecoverTaskF func(*drivers.TaskHandle) error
|
|
StartTaskF func(*drivers.TaskConfig) (*drivers.TaskHandle, *drivers.DriverNetwork, error)
|
|
WaitTaskF func(context.Context, string) (<-chan *drivers.ExitResult, error)
|
|
StopTaskF func(string, time.Duration, string) error
|
|
DestroyTaskF func(string, bool) error
|
|
InspectTaskF func(string) (*drivers.TaskStatus, error)
|
|
TaskStatsF func(context.Context, string, time.Duration) (<-chan *drivers.TaskResourceUsage, error)
|
|
TaskEventsF func(context.Context) (<-chan *drivers.TaskEvent, error)
|
|
SignalTaskF func(string, string) error
|
|
ExecTaskF func(string, []string, time.Duration) (*drivers.ExecTaskResult, error)
|
|
ExecTaskStreamingF func(context.Context, string, *drivers.ExecOptions) (*drivers.ExitResult, error)
|
|
MockNetworkManager
|
|
}
|
|
|
|
type MockNetworkManager struct {
|
|
CreateNetworkF func(string, *drivers.NetworkCreateRequest) (*drivers.NetworkIsolationSpec, bool, error)
|
|
DestroyNetworkF func(string, *drivers.NetworkIsolationSpec) error
|
|
}
|
|
|
|
func (m *MockNetworkManager) CreateNetwork(allocID string, req *drivers.NetworkCreateRequest) (*drivers.NetworkIsolationSpec, bool, error) {
|
|
return m.CreateNetworkF(allocID, req)
|
|
}
|
|
func (m *MockNetworkManager) DestroyNetwork(id string, spec *drivers.NetworkIsolationSpec) error {
|
|
return m.DestroyNetworkF(id, spec)
|
|
}
|
|
|
|
func (d *MockDriver) TaskConfigSchema() (*hclspec.Spec, error) { return d.TaskConfigSchemaF() }
|
|
func (d *MockDriver) Fingerprint(ctx context.Context) (<-chan *drivers.Fingerprint, error) {
|
|
return d.FingerprintF(ctx)
|
|
}
|
|
func (d *MockDriver) Capabilities() (*drivers.Capabilities, error) { return d.CapabilitiesF() }
|
|
func (d *MockDriver) RecoverTask(h *drivers.TaskHandle) error { return d.RecoverTaskF(h) }
|
|
func (d *MockDriver) StartTask(c *drivers.TaskConfig) (*drivers.TaskHandle, *drivers.DriverNetwork, error) {
|
|
return d.StartTaskF(c)
|
|
}
|
|
func (d *MockDriver) WaitTask(ctx context.Context, id string) (<-chan *drivers.ExitResult, error) {
|
|
return d.WaitTaskF(ctx, id)
|
|
}
|
|
func (d *MockDriver) StopTask(taskID string, timeout time.Duration, signal string) error {
|
|
return d.StopTaskF(taskID, timeout, signal)
|
|
}
|
|
func (d *MockDriver) DestroyTask(taskID string, force bool) error {
|
|
return d.DestroyTaskF(taskID, force)
|
|
}
|
|
func (d *MockDriver) InspectTask(taskID string) (*drivers.TaskStatus, error) {
|
|
return d.InspectTaskF(taskID)
|
|
}
|
|
func (d *MockDriver) TaskStats(ctx context.Context, taskID string, i time.Duration) (<-chan *drivers.TaskResourceUsage, error) {
|
|
return d.TaskStatsF(ctx, taskID, i)
|
|
}
|
|
func (d *MockDriver) TaskEvents(ctx context.Context) (<-chan *drivers.TaskEvent, error) {
|
|
return d.TaskEventsF(ctx)
|
|
}
|
|
func (d *MockDriver) SignalTask(taskID string, signal string) error {
|
|
return d.SignalTaskF(taskID, signal)
|
|
}
|
|
func (d *MockDriver) ExecTask(taskID string, cmd []string, timeout time.Duration) (*drivers.ExecTaskResult, error) {
|
|
return d.ExecTaskF(taskID, cmd, timeout)
|
|
}
|
|
|
|
func (d *MockDriver) ExecTaskStreaming(ctx context.Context, taskID string, execOpts *drivers.ExecOptions) (*drivers.ExitResult, error) {
|
|
return d.ExecTaskStreamingF(ctx, taskID, execOpts)
|
|
}
|
|
|
|
// SetEnvvars sets path and host env vars depending on the FS isolation used.
|
|
func SetEnvvars(envBuilder *taskenv.Builder, fsi drivers.FSIsolation, taskDir *allocdir.TaskDir, conf *config.Config) {
|
|
|
|
envBuilder.SetClientTaskRoot(taskDir.Dir)
|
|
envBuilder.SetClientSharedAllocDir(taskDir.SharedAllocDir)
|
|
envBuilder.SetClientTaskLocalDir(taskDir.LocalDir)
|
|
envBuilder.SetClientTaskSecretsDir(taskDir.SecretsDir)
|
|
|
|
// Set driver-specific environment variables
|
|
switch fsi {
|
|
case drivers.FSIsolationNone:
|
|
// Use host paths
|
|
envBuilder.SetAllocDir(taskDir.SharedAllocDir)
|
|
envBuilder.SetTaskLocalDir(taskDir.LocalDir)
|
|
envBuilder.SetSecretsDir(taskDir.SecretsDir)
|
|
default:
|
|
// filesystem isolation; use container paths
|
|
envBuilder.SetAllocDir(allocdir.SharedAllocContainerPath)
|
|
envBuilder.SetTaskLocalDir(allocdir.TaskLocalContainerPath)
|
|
envBuilder.SetSecretsDir(allocdir.TaskSecretsContainerPath)
|
|
}
|
|
|
|
// Set the host environment variables for non-image based drivers
|
|
if fsi != drivers.FSIsolationImage {
|
|
// COMPAT(1.0) using inclusive language, blacklist is kept for backward compatibility.
|
|
filter := strings.Split(conf.ReadAlternativeDefault(
|
|
[]string{"env.denylist", "env.blacklist"},
|
|
config.DefaultEnvDenylist,
|
|
), ",")
|
|
envBuilder.SetHostEnvvars(filter)
|
|
}
|
|
}
|