drivers: set -1 exit code in case executor gets killed (#25453)

Nomad driver handles incorrectly set exit code 0 in case of executor failure.
This corrects that behavior.

---------

Co-authored-by: Tim Gross <tgross@hashicorp.com>

.changelog/25453.txt

@@ -0,0 +1,3 @@
+```release-note:bug
+drivers: set -1 exit code in case of executor failure for the exec, raw_exec, java, and qemu task drivers
+```

drivers/exec/driver.go

@@ -595,6 +595,12 @@ func (d *Driver) handleWait(ctx context.Context, handle *taskHandle, ch chan *dr
 		result = &drivers.ExitResult{
 			Err: fmt.Errorf("executor: error waiting on process: %v", err),
 		}
+		// if process state is nil, we've probably been killed, so return a reasonable
+		// exit state to the handlers
+		if ps == nil {
+			result.ExitCode = -1
+			result.OOMKilled = false
+		}
 	} else {
 		result = &drivers.ExitResult{
 			ExitCode:  ps.ExitCode,

drivers/java/driver.go

@@ -604,6 +604,12 @@ func (d *Driver) handleWait(ctx context.Context, handle *taskHandle, ch chan *dr
 		result = &drivers.ExitResult{
 			Err: fmt.Errorf("executor: error waiting on process: %v", err),
 		}
+		// if process state is nil, we've probably been killed, so return a reasonable
+		// exit state to the handlers
+		if ps == nil {
+			result.ExitCode = -1
+			result.OOMKilled = false
+		}
 	} else {
 		result = &drivers.ExitResult{
 			ExitCode:  ps.ExitCode,

drivers/qemu/driver.go

@@ -743,6 +743,12 @@ func (d *Driver) handleWait(ctx context.Context, handle *taskHandle, ch chan *dr
 		result = &drivers.ExitResult{
 			Err: fmt.Errorf("executor: error waiting on process: %v", err),
 		}
+		// if process state is nil, we've probably been killed, so return a reasonable
+		// exit state to the handlers
+		if ps == nil {
+			result.ExitCode = -1
+			result.OOMKilled = false
+		}
 	} else {
 		result = &drivers.ExitResult{
 			ExitCode:  ps.ExitCode,

drivers/rawexec/driver.go

@@ -471,6 +471,12 @@ func (d *Driver) handleWait(ctx context.Context, handle *taskHandle, ch chan *dr
 		result = &drivers.ExitResult{
 			Err: fmt.Errorf("executor: error waiting on process: %v", err),
 		}
+		// if process state is nil, we've probably been killed, so return a reasonable
+		// exit state to the handlers
+		if ps == nil {
+			result.ExitCode = -1
+			result.OOMKilled = false
+		}
 	} else {
 		result = &drivers.ExitResult{
 			ExitCode:  ps.ExitCode,

drivers/rawexec/driver_unix_test.go

@@ -580,3 +580,59 @@ func TestRawExec_Validate(t *testing.T) {
 		must.Eq(t, tc.exp, d.Validate(tc.driverConfig))
 	}
 }
+
+func TestRawExecDriver_ExecutorKilled_ExitCode(t *testing.T) {
+	ci.Parallel(t)
+	clienttestutil.ExecCompatible(t)
+
+	d := newEnabledRawExecDriver(t)
+	harness := dtestutil.NewDriverHarness(t, d)
+	defer harness.Kill()
+
+	allocID := uuid.Generate()
+	taskName := "sleep"
+	task := &drivers.TaskConfig{
+		AllocID:   allocID,
+		ID:        uuid.Generate(),
+		Name:      taskName,
+		Env:       defaultEnv(),
+		Resources: testResources(allocID, taskName),
+	}
+
+	cleanup := harness.MkAllocDir(task, false)
+	defer cleanup()
+
+	tc := &TaskConfig{
+		Command: testtask.Path(),
+		Args:    []string{"sleep", "10s"},
+	}
+	must.NoError(t, task.EncodeConcreteDriverConfig(&tc))
+	testtask.SetTaskConfigEnv(task)
+
+	harness.MakeTaskCgroup(allocID, taskName)
+	handle, _, err := harness.StartTask(task)
+	must.NoError(t, err)
+
+	// Decode driver state to get executor PID
+	var driverState TaskState
+	must.NoError(t, handle.GetDriverState(&driverState))
+
+	// Kill the executor and wait until it's gone
+	pid := driverState.ReattachConfig.Pid
+	must.NoError(t, err)
+	must.NoError(t, syscall.Kill(pid, syscall.SIGKILL))
+
+	// Make sure the right exit code is set
+	waitCh, err := harness.WaitTask(context.Background(), task.ID)
+	must.NoError(t, err)
+	select {
+	case res := <-waitCh:
+		must.False(t, res.Successful())
+		must.Eq(t, -1, res.ExitCode)
+		must.Eq(t, false, res.OOMKilled)
+	case <-time.After(10 * time.Second):
+		must.Unreachable(t, must.Sprint("exceeded wait timeout"))
+	}
+
+	must.NoError(t, harness.DestroyTask(task.ID, true))
+}