Implement snapshot restore

command/commands.go

@@ -517,6 +517,11 @@ func Commands(metaPtr *Meta, agentUi cli.Ui) map[string]cli.CommandFactory {
 				Meta: meta,
 			}, nil
 		},
+		"operator snapshot restore": func() (cli.Command, error) {
+			return &OperatorSnapshotRestoreCommand{
+				Meta: meta,
+			}, nil
+		},
 
 		"plan": func() (cli.Command, error) {
 			return &JobPlanCommand{

command/operator_snapshot_inspect_test.go

@@ -6,6 +6,7 @@ import (
 	"path/filepath"
 	"testing"
 
+	"github.com/hashicorp/nomad/api"
 	"github.com/hashicorp/nomad/command/agent"
 	"github.com/mitchellh/cli"
 	"github.com/stretchr/testify/require"
@@ -14,7 +15,7 @@ import (
 func TestOperatorSnapshotInspect_Works(t *testing.T) {
 	t.Parallel()
 
-	snapPath := generateSnapshotFile(t)
+	snapPath := generateSnapshotFile(t, nil)
 
 	ui := new(cli.MockUi)
 	cmd := &OperatorSnapshotInspectCommand{Meta: Meta{Ui: ui}}
@@ -67,24 +68,29 @@ func TestOperatorSnapshotInspect_HandlesFailure(t *testing.T) {
 
 }
 
-func generateSnapshotFile(t *testing.T) string {
+func generateSnapshotFile(t *testing.T, prepare func(srv *agent.TestAgent, client *api.Client, url string)) string {
 
 	tmpDir, err := ioutil.TempDir("", "nomad-tempdir")
 	require.NoError(t, err)
 
 	t.Cleanup(func() { os.RemoveAll(tmpDir) })
 
-	srv, _, url := testServer(t, false, func(c *agent.Config) {
+	srv, api, url := testServer(t, false, func(c *agent.Config) {
 		c.DevMode = false
 		c.DataDir = filepath.Join(tmpDir, "server")
 
+		c.Client.Enabled = false
 		c.AdvertiseAddrs.HTTP = "127.0.0.1"
 		c.AdvertiseAddrs.RPC = "127.0.0.1"
 		c.AdvertiseAddrs.Serf = "127.0.0.1"
 	})
 
 	defer srv.Shutdown()
 
+	if prepare != nil {
+		prepare(srv, api, url)
+	}
+
 	ui := new(cli.MockUi)
 	cmd := &OperatorSnapshotSaveCommand{Meta: Meta{Ui: ui}}
 

command/operator_snapshot_restore.go

@@ -0,0 +1,95 @@
+package command
+
+import (
+	"fmt"
+	"os"
+	"strings"
+
+	"github.com/hashicorp/nomad/api"
+	"github.com/posener/complete"
+)
+
+type OperatorSnapshotRestoreCommand struct {
+	Meta
+}
+
+func (c *OperatorSnapshotRestoreCommand) Help() string {
+	helpText := `
+Usage: nomad snapshot restore [options] FILE
+
+  Restores an atomic, point-in-time snapshot of the state of the Nomad servers
+  which includes jobs, nodes, allocations, periodic jobs, and ACLs.
+
+  Restores involve a potentially dangerous low-level Raft operation that is not
+  designed to handle server failures during a restore. This command is primarily
+  intended to be used when recovering from a disaster, restoring into a fresh
+  cluster of Nomad servers.
+
+  If ACLs are enabled, a management token must be supplied in order to perform
+  snapshot operations.
+
+  To restore a snapshot from the file "backup.snap":
+
+    $ nomad snapshot restore backup.snap
+
+General Options:
+
+  ` + generalOptionsUsage()
+	return strings.TrimSpace(helpText)
+}
+
+func (c *OperatorSnapshotRestoreCommand) AutocompleteFlags() complete.Flags {
+	return c.Meta.AutocompleteFlags(FlagSetClient)
+}
+
+func (c *OperatorSnapshotRestoreCommand) AutocompleteArgs() complete.Predictor {
+	return complete.PredictNothing
+}
+
+func (c *OperatorSnapshotRestoreCommand) Synopsis() string {
+	return "Restore snapshot of Nomad server state"
+}
+
+func (c *OperatorSnapshotRestoreCommand) Name() string { return "operator snapshot restore" }
+
+func (c *OperatorSnapshotRestoreCommand) Run(args []string) int {
+	flags := c.Meta.FlagSet(c.Name(), FlagSetClient)
+	flags.Usage = func() { c.Ui.Output(c.Help()) }
+
+	if err := flags.Parse(args); err != nil {
+		c.Ui.Error(fmt.Sprintf("Failed to parse args: %v", err))
+		return 1
+	}
+
+	// Check for misuse
+	args = flags.Args()
+	if len(args) != 1 {
+		c.Ui.Error("This command takes one: <filename>")
+		c.Ui.Error(commandErrorText(c))
+		return 1
+	}
+
+	snap, err := os.Open(args[0])
+	if err != nil {
+		c.Ui.Error(fmt.Sprintf("Error opening snapshot file: %q", err))
+		return 1
+	}
+	defer snap.Close()
+
+	// Set up a client.
+	client, err := c.Meta.Client()
+	if err != nil {
+		c.Ui.Error(fmt.Sprintf("Error initializing client: %s", err))
+		return 1
+	}
+
+	// Fetch the current configuration.
+	_, err = client.Operator().SnapshotRestore(snap, &api.WriteOptions{})
+	if err != nil {
+		c.Ui.Error(fmt.Sprintf("Failed to get restore snapshot: %v", err))
+		return 1
+	}
+
+	c.Ui.Output("Snapshot Restored")
+	return 0
+}

command/operator_snapshot_restore_test.go

@@ -0,0 +1,96 @@
+package command
+
+import (
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/hashicorp/nomad/api"
+	"github.com/hashicorp/nomad/command/agent"
+	"github.com/hashicorp/nomad/nomad/structs"
+	"github.com/mitchellh/cli"
+	"github.com/stretchr/testify/require"
+)
+
+func TestOperatorSnapshotRestore_Works(t *testing.T) {
+	t.Parallel()
+
+	tmpDir, err := ioutil.TempDir("", "nomad-tempdir")
+	require.NoError(t, err)
+	defer os.RemoveAll(tmpDir)
+
+	snapshotPath := generateSnapshotFile(t, func(srv *agent.TestAgent, client *api.Client, url string) {
+		sampleJob := `
+job "snapshot-test-job" {
+	type = "service"
+	datacenters = [ "dc1" ]
+	group "group1" {
+		count = 1
+		task "task1" {
+			driver = "exec"
+			resources = {
+				cpu = 1000
+				memory = 512
+			}
+		}
+	}
+
+}`
+
+		ui := new(cli.MockUi)
+		cmd := &JobRunCommand{Meta: Meta{Ui: ui}}
+		cmd.JobGetter.testStdin = strings.NewReader(sampleJob)
+
+		code := cmd.Run([]string{"--address=" + url, "-detach", "-"})
+		require.Zero(t, code)
+	})
+
+	srv, _, url := testServer(t, false, func(c *agent.Config) {
+		c.DevMode = false
+		c.DataDir = filepath.Join(tmpDir, "server1")
+
+		c.Client.Enabled = false
+		c.AdvertiseAddrs.HTTP = "127.0.0.1"
+		c.AdvertiseAddrs.RPC = "127.0.0.1"
+		c.AdvertiseAddrs.Serf = "127.0.0.1"
+	})
+
+	defer srv.Shutdown()
+
+	// job is not found before restore
+	j, err := srv.Agent.Server().State().JobByID(nil, structs.DefaultNamespace, "snapshot-test-job")
+	require.NoError(t, err)
+	require.Nil(t, j)
+
+	ui := new(cli.MockUi)
+	cmd := &OperatorSnapshotRestoreCommand{Meta: Meta{Ui: ui}}
+
+	code := cmd.Run([]string{"--address=" + url, snapshotPath})
+	require.Empty(t, ui.ErrorWriter.String())
+	require.Zero(t, code)
+	require.Contains(t, ui.OutputWriter.String(), "Snapshot Restored")
+
+	foundJob, err := srv.Agent.Server().State().JobByID(nil, structs.DefaultNamespace, "snapshot-test-job")
+	require.NoError(t, err)
+	require.Equal(t, "snapshot-test-job", foundJob.ID)
+}
+
+func TestOperatorSnapshotRestore_Fails(t *testing.T) {
+	t.Parallel()
+
+	ui := new(cli.MockUi)
+	cmd := &OperatorSnapshotRestoreCommand{Meta: Meta{Ui: ui}}
+
+	// Fails on misuse
+	code := cmd.Run([]string{"some", "bad", "args"})
+	require.Equal(t, 1, code)
+	require.Contains(t, ui.ErrorWriter.String(), commandErrorText(cmd))
+	ui.ErrorWriter.Reset()
+
+	// Fails when specified file does not exist
+	code = cmd.Run([]string{"/unicorns/leprechauns"})
+	require.Equal(t, 1, code)
+	require.Contains(t, ui.ErrorWriter.String(), "no such file")
+}

command/operator_snapshot_save_test.go

@@ -49,3 +49,21 @@ func TestOperatorSnapshotSave_Works(t *testing.T) {
 	require.NoError(t, err)
 	require.NotZero(t, meta.Index)
 }
+
+func TestOperatorSnapshotSave_Fails(t *testing.T) {
+	t.Parallel()
+
+	ui := new(cli.MockUi)
+	cmd := &OperatorSnapshotSaveCommand{Meta: Meta{Ui: ui}}
+
+	// Fails on misuse
+	code := cmd.Run([]string{"some", "bad", "args"})
+	require.Equal(t, 1, code)
+	require.Contains(t, ui.ErrorWriter.String(), commandErrorText(cmd))
+	ui.ErrorWriter.Reset()
+
+	// Fails when specified file does not exist
+	code = cmd.Run([]string{"/unicorns/leprechauns"})
+	require.Equal(t, 1, code)
+	require.Contains(t, ui.ErrorWriter.String(), "no such file")
+}