Merge branch 'master' into f-bolt-db

.travis.yml

@@ -7,7 +7,7 @@ services:
 language: go
 
 go:
-  -  1.8
+  -  1.8.x
 
 branches:
   only:

CHANGELOG.md

@@ -4,10 +4,14 @@ IMPROVEMENTS:
  * core: Back-pressure when evaluations are nacked and ensure scheduling
    progress on evaluation failures [GH-2555]
  * core: Track multiple job versions and add a stopped state for jobs [GH-2566]
+ * api: Add `verify_https_client` to require certificates from HTTP clients
+   [GH-2587]
  * api/job: Ability to revert job to older versions [GH-2575]
  * client: Fingerprint all routable addresses on an interface including IPv6
    addresses [GH-2536]
  * client: Hash host ID so its stable and well distributed [GH-2541]
+ * client: Environment variables for client DC and Region [GH-2507]
+ * config: Support Unix socket addresses for Consul [GH-2622]
  * driver/docker: Allow specifying extra hosts [GH-2547]
  * driver/docker: Allow setting container IP with user defined networks
    [GH-2535]
@@ -22,6 +26,7 @@ BUG FIXES:
  * client/artifact: Handle tars where file in directory is listed before
    directory [GH-2524]
  * driver/exec: Properly set file/dir ownership in chroots [GH-2552]
+ * driver/docker: Fix panic in Docker driver on Windows [GH-2614]
  * server: Reject non-TLS clients when TLS enabled [GH-2525]
  * server: Fix a panic in plan evaluation with partial failures and all_at_once
    set [GH-2544]

Vagrantfile

@@ -20,7 +20,8 @@ sudo DEBIAN_FRONTEND=noninteractive apt-get install -y build-essential curl git-
      liblxc1 lxc-dev lxc-templates                      \
      gcc-5-aarch64-linux-gnu binutils-aarch64-linux-gnu \
      libc6-dev-i386 linux-libc-dev:i386                 \
-     gcc-5-arm-linux-gnueabihf gcc-5-multilib-arm-linux-gnueabihf binutils-arm-linux-gnueabihf
+     gcc-5-arm-linux-gnueabihf gcc-5-multilib-arm-linux-gnueabihf binutils-arm-linux-gnueabihf \
+     gcc-mingw-w64 binutils-mingw-w64
 
 # Setup go, for development of Nomad
 SRCROOT="/opt/go"

client/alloc_runner.go

@@ -276,8 +276,8 @@ func (r *AllocRunner) RestoreState() error {
 			continue
 		}
 
-		if err := tr.RestoreState(); err != nil {
-			r.logger.Printf("[ERR] client: failed to restore state for alloc %s task '%s': %v", r.alloc.ID, name, err)
+		if restartReason, err := tr.RestoreState(); err != nil {
+			r.logger.Printf("[ERR] client: failed to restore state for alloc %s task %q: %v", r.alloc.ID, name, err)
 			mErr.Errors = append(mErr.Errors, err)
 		} else if !r.alloc.TerminalStatus() {
 			// Only start if the alloc isn't in a terminal status.
@@ -288,6 +288,11 @@ func (r *AllocRunner) RestoreState() error {
 					r.logger.Printf("[WARN] client: initial save state for alloc %s task %s failed: %v", r.alloc.ID, name, err)
 				}
 			}
+
+			// Restart task runner if RestoreState gave a reason
+			if restartReason != "" {
+				tr.Restart("upgrade", restartReason)
+			}
 		}
 	}
 

client/alloc_runner_test.go

@@ -34,6 +34,7 @@ func (m *MockAllocStateUpdater) Update(alloc *structs.Allocation) {
 func testAllocRunnerFromAlloc(alloc *structs.Allocation, restarts bool) (*MockAllocStateUpdater, *AllocRunner) {
 	logger := testLogger()
 	conf := config.DefaultConfig()
+	conf.Node = mock.Node()
 	conf.StateDir = os.TempDir()
 	conf.AllocDir = os.TempDir()
 	tmp, _ := ioutil.TempFile("", "state-db")
@@ -502,6 +503,84 @@ func TestAllocRunner_SaveRestoreState_TerminalAlloc(t *testing.T) {
 	})
 }
 
+// TestAllocRunner_SaveRestoreState_Upgrade asserts that pre-0.6 exec tasks are
+// restarted on upgrade.
+func TestAllocRunner_SaveRestoreState_Upgrade(t *testing.T) {
+	alloc := mock.Alloc()
+	task := alloc.Job.TaskGroups[0].Tasks[0]
+	task.Driver = "mock_driver"
+	task.Config = map[string]interface{}{
+		"exit_code": "0",
+		"run_for":   "10s",
+	}
+
+	upd, ar := testAllocRunnerFromAlloc(alloc, false)
+	// Hack in old version to cause an upgrade on RestoreState
+	origConfig := ar.config.Copy()
+	ar.config.Version = "0.5.6"
+	go ar.Run()
+
+	// Snapshot state
+	testutil.WaitForResult(func() (bool, error) {
+		return len(ar.tasks) == 1, nil
+	}, func(err error) {
+		t.Fatalf("task never started: %v", err)
+	})
+
+	err := ar.SaveState()
+	if err != nil {
+		t.Fatalf("err: %v", err)
+	}
+
+	// Create a new alloc runner
+	l2 := prefixedTestLogger("----- ar2:  ")
+	ar2 := NewAllocRunner(l2, origConfig, upd.Update,
+		&structs.Allocation{ID: ar.alloc.ID}, ar.vaultClient,
+		ar.consulClient)
+	err = ar2.RestoreState()
+	if err != nil {
+		t.Fatalf("err: %v", err)
+	}
+	go ar2.Run()
+
+	testutil.WaitForResult(func() (bool, error) {
+		if len(ar2.tasks) != 1 {
+			return false, fmt.Errorf("Incorrect number of tasks")
+		}
+
+		if upd.Count < 3 {
+			return false, nil
+		}
+
+		for _, ev := range ar2.alloc.TaskStates["web"].Events {
+			if strings.HasSuffix(ev.RestartReason, pre06ScriptCheckReason) {
+				return true, nil
+			}
+		}
+		return false, fmt.Errorf("no restart with proper reason found")
+	}, func(err error) {
+		t.Fatalf("err: %v\nAllocs: %#v\nWeb State: %#v", err, upd.Allocs, ar2.alloc.TaskStates["web"])
+	})
+
+	// Destroy and wait
+	ar2.Destroy()
+	start := time.Now()
+
+	testutil.WaitForResult(func() (bool, error) {
+		alloc := ar2.Alloc()
+		if alloc.ClientStatus != structs.AllocClientStatusComplete {
+			return false, fmt.Errorf("Bad client status; got %v; want %v", alloc.ClientStatus, structs.AllocClientStatusComplete)
+		}
+		return true, nil
+	}, func(err error) {
+		t.Fatalf("err: %v %#v %#v", err, upd.Allocs[0], ar.alloc.TaskStates)
+	})
+
+	if time.Since(start) > time.Duration(testutil.TestMultiplier()*5)*time.Second {
+		t.Fatalf("took too long to terminate")
+	}
+}
+
 // Ensure pre-#2132 state files containing the Context struct are properly
 // migrated to the new format.
 //

client/consul_template_test.go

@@ -121,7 +121,10 @@ func newTestHarness(t *testing.T, templates []*structs.Template, consul, vault b
 	harness.taskDir = d
 
 	if consul {
-		harness.consul = ctestutil.NewTestServer(t)
+		harness.consul, err = ctestutil.NewTestServer()
+		if err != nil {
+			t.Fatalf("error starting test Consul server: %v", err)
+		}
 		harness.config.ConsulConfig = &sconfig.ConsulConfig{
 			Addr: harness.consul.HTTPAddr,
 		}
@@ -445,7 +448,7 @@ func TestTaskTemplateManager_Unblock_Consul(t *testing.T) {
 	}
 
 	// Write the key to Consul
-	harness.consul.SetKV(key, []byte(content))
+	harness.consul.SetKV(t, key, []byte(content))
 
 	// Wait for the unblock
 	select {
@@ -563,7 +566,7 @@ func TestTaskTemplateManager_Unblock_Multi_Template(t *testing.T) {
 	}
 
 	// Write the key to Consul
-	harness.consul.SetKV(consulKey, []byte(consulContent))
+	harness.consul.SetKV(t, consulKey, []byte(consulContent))
 
 	// Wait for the unblock
 	select {
@@ -612,7 +615,7 @@ func TestTaskTemplateManager_Rerender_Noop(t *testing.T) {
 	}
 
 	// Write the key to Consul
-	harness.consul.SetKV(key, []byte(content1))
+	harness.consul.SetKV(t, key, []byte(content1))
 
 	// Wait for the unblock
 	select {
@@ -633,7 +636,7 @@ func TestTaskTemplateManager_Rerender_Noop(t *testing.T) {
 	}
 
 	// Update the key in Consul
-	harness.consul.SetKV(key, []byte(content2))
+	harness.consul.SetKV(t, key, []byte(content2))
 
 	select {
 	case <-harness.mockHooks.RestartCh:
@@ -697,8 +700,8 @@ func TestTaskTemplateManager_Rerender_Signal(t *testing.T) {
 	}
 
 	// Write the key to Consul
-	harness.consul.SetKV(key1, []byte(content1_1))
-	harness.consul.SetKV(key2, []byte(content2_1))
+	harness.consul.SetKV(t, key1, []byte(content1_1))
+	harness.consul.SetKV(t, key2, []byte(content2_1))
 
 	// Wait for the unblock
 	select {
@@ -712,8 +715,8 @@ func TestTaskTemplateManager_Rerender_Signal(t *testing.T) {
 	}
 
 	// Update the keys in Consul
-	harness.consul.SetKV(key1, []byte(content1_2))
-	harness.consul.SetKV(key2, []byte(content2_2))
+	harness.consul.SetKV(t, key1, []byte(content1_2))
+	harness.consul.SetKV(t, key2, []byte(content2_2))
 
 	// Wait for signals
 	timeout := time.After(time.Duration(1*testutil.TestMultiplier()) * time.Second)
@@ -782,7 +785,7 @@ func TestTaskTemplateManager_Rerender_Restart(t *testing.T) {
 	}
 
 	// Write the key to Consul
-	harness.consul.SetKV(key1, []byte(content1_1))
+	harness.consul.SetKV(t, key1, []byte(content1_1))
 
 	// Wait for the unblock
 	select {
@@ -792,7 +795,7 @@ func TestTaskTemplateManager_Rerender_Restart(t *testing.T) {
 	}
 
 	// Update the keys in Consul
-	harness.consul.SetKV(key1, []byte(content1_2))
+	harness.consul.SetKV(t, key1, []byte(content1_2))
 
 	// Wait for restart
 	timeout := time.After(time.Duration(1*testutil.TestMultiplier()) * time.Second)
@@ -878,7 +881,7 @@ func TestTaskTemplateManager_Signal_Error(t *testing.T) {
 	harness.mockHooks.SignalError = fmt.Errorf("test error")
 
 	// Write the key to Consul
-	harness.consul.SetKV(key1, []byte(content1))
+	harness.consul.SetKV(t, key1, []byte(content1))
 
 	// Wait a little
 	select {
@@ -888,7 +891,7 @@ func TestTaskTemplateManager_Signal_Error(t *testing.T) {
 	}
 
 	// Write the key to Consul
-	harness.consul.SetKV(key1, []byte(content2))
+	harness.consul.SetKV(t, key1, []byte(content2))
 
 	// Wait for kill channel
 	select {

client/driver/driver.go

@@ -309,6 +309,8 @@ func GetTaskEnv(taskDir *allocdir.TaskDir, node *structs.Node,
 	env := env.NewTaskEnvironment(node).
 		SetTaskMeta(alloc.Job.CombinedTaskMeta(alloc.TaskGroup, task.Name)).
 		SetJobName(alloc.Job.Name).
+		SetDatacenterName(node.Datacenter).
+		SetRegionName(conf.Region).
 		SetEnvvars(task.Env).
 		SetTaskName(task.Name)
 

client/driver/driver_test.go

@@ -74,6 +74,8 @@ func testConfig() *config.Config {
 	conf.StateDir = os.TempDir()
 	conf.AllocDir = os.TempDir()
 	conf.MaxKillTimeout = 10 * time.Second
+	conf.Region = "global"
+	conf.Node = mock.Node()
 	return conf
 }
 
@@ -88,6 +90,7 @@ type testContext struct {
 // It is up to the caller to call AllocDir.Destroy to cleanup.
 func testDriverContexts(t *testing.T, task *structs.Task) *testContext {
 	cfg := testConfig()
+	cfg.Node = mock.Node()
 	allocDir := allocdir.NewAllocDir(testLogger(), filepath.Join(cfg.AllocDir, structs.GenerateUUID()))
 	if err := allocDir.Build(); err != nil {
 		t.Fatalf("AllocDir.Build() failed: %v", err)
@@ -162,7 +165,7 @@ func setupTaskEnv(t *testing.T, driver string) (*allocdir.TaskDir, map[string]st
 	conf := testConfig()
 	allocDir := allocdir.NewAllocDir(testLogger(), filepath.Join(conf.AllocDir, alloc.ID))
 	taskDir := allocDir.NewTaskDir(task.Name)
-	env, err := GetTaskEnv(taskDir, nil, task, alloc, conf, "")
+	env, err := GetTaskEnv(taskDir, conf.Node, task, alloc, conf, "")
 	if err != nil {
 		t.Fatalf("GetTaskEnv() failed: %v", err)
 	}
@@ -209,6 +212,8 @@ func setupTaskEnv(t *testing.T, driver string) (*allocdir.TaskDir, map[string]st
 		"NOMAD_ALLOC_NAME":              alloc.Name,
 		"NOMAD_TASK_NAME":               task.Name,
 		"NOMAD_JOB_NAME":                alloc.Job.Name,
+		"NOMAD_DC":                      "dc1",
+		"NOMAD_REGION":                  "global",
 	}
 
 	act := env.EnvMap()