Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Handle specific exception codes on RCI call. #4457

Open
wants to merge 1 commit into
base: dev
Choose a base branch
from
Open

Handle specific exception codes on RCI call. #4457

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 14 additions & 3 deletions agent/app/agent.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -447,10 +447,15 @@ func (agent *ecsAgent) doStart(containerChangeEventStream *eventstream.EventStre
// Register the container instance
err = agent.registerContainerInstance(client, vpcSubnetAttributes)
if err != nil {
if isTransient(err) {
if isTerminal(err) {
// On unrecoverable error codes, agent should terminally exit.
seelog.Warnf("Agent will terminally exit, unable to register container instance: %v", err)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: should use logger instead of seelog here?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

seelog.Warnf is chosen here to conform to existing style of using it for warning messages in this specific file i.e. there were 9 seelog.warn vs 2 logger.warn - so I opted for the more commonly used

There are 9 cases I used as reference found :

  1. amazon-ecs-agent/agent/app/agent.go

    Line 945 in 41d593c

    seelog.Warnf("Error initializing metrics engine: %v", err)

  2. amazon-ecs-agent/agent/app/agent.go

    Line 953 in 41d593c

    seelog.Warnf("Error creating telemetry session: %v", err)

Where as logger is also used but mostly for logger.critical and logger.info. There are only 2 cases of logger.Warn:

amazon-ecs-agent/agent/app/agent.go

Line 1036 in 41d593c

logger.Warn("Failed to get docker version from task engine", logger.Fields{

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Got it, thanks for the justification. I am no expert but it seems more idiomatic to conform to the in-house logger used in this library

return exitcodes.ExitTerminal
} else if isTransient(err) {
return exitcodes.ExitError
}
return exitcodes.ExitTerminal
// Other errors are considered recoverable and will be retried.
return exitcodes.ExitError
}

// Load Managed Daemon images asynchronously
Expand Down Expand Up @@ -840,7 +845,13 @@ func (agent *ecsAgent) registerContainerInstance(
logger.Critical("Instance registration attempt with an invalid parameter", logger.Fields{
field.Error: err,
})
return err
return terminalError{err}
}
if utils.IsAWSErrorCodeEqual(err, ecsmodel.ErrCodeClientException) {
logger.Critical("Instance registration attempt with client performing invalid action", logger.Fields{
field.Error: err,
})
return terminalError{err}
}
if _, ok := err.(apierrors.AttributeError); ok {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it possible this is also terminal error?

Instance registration attempt with invalid attribute(s)

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Handling that specific case is beyond scope of this change.

However, invalidParameterException is handled above, I would not expect attributeError to fire because instance attributes are an instance parameter of the RCI call. Ecs agent should correct me here if I'm wrong

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sounds good to me. We can focus on the cases we care about now.

attributeErrorMsg := ""
Expand Down
129 changes: 127 additions & 2 deletions agent/app/agent_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -222,7 +222,7 @@ func TestDoStartNewTaskEngineError(t *testing.T) {
assert.Equal(t, exitcodes.ExitTerminal, exitCode)
}

func TestDoStartRegisterContainerInstanceErrorTerminal(t *testing.T) {
func TestDoStartRegisterContainerInstanceErrorTransient(t *testing.T) {
ctrl, credentialsManager, state, imageManager, client,
dockerClient, _, _, execCmdMgr, _ := setup(t)
defer ctrl.Finish()
Expand Down Expand Up @@ -282,7 +282,7 @@ func TestDoStartRegisterContainerInstanceErrorTerminal(t *testing.T) {

exitCode := agent.doStart(eventstream.NewEventStream("events", ctx),
credentialsManager, state, imageManager, client, execCmdMgr)
assert.Equal(t, exitcodes.ExitTerminal, exitCode)
assert.Equal(t, exitcodes.ExitError, exitCode)
}

func TestDoStartRegisterContainerInstanceErrorNonTerminal(t *testing.T) {
Expand Down Expand Up @@ -1499,6 +1499,131 @@ func TestRegisterContainerInstanceInvalidParameterTerminalError(t *testing.T) {
credentialsManager, state, imageManager, client, execCmdMgr)
assert.Equal(t, exitcodes.ExitTerminal, exitCode)
}

func TestRegisterContainerInstanceInvalidClientError(t *testing.T) {
ctrl, credentialsManager, state, imageManager, client,
dockerClient, _, _, execCmdMgr, _ := setup(t)
defer ctrl.Finish()

mockCredentialsProvider := app_mocks.NewMockCredentialsProvider(ctrl)
mockMobyPlugins := mock_mobypkgwrapper.NewMockPlugins(ctrl)
mockEC2Metadata := mock_ec2.NewMockEC2MetadataClient(ctrl)
mockPauseLoader := mock_loader.NewMockLoader(ctrl)

mockPauseLoader.EXPECT().IsLoaded(gomock.Any()).Return(false, nil).AnyTimes()
mockPauseLoader.EXPECT().LoadImage(gomock.Any(), gomock.Any(), gomock.Any()).Return(nil, nil).AnyTimes()
mockEC2Metadata.EXPECT().PrimaryENIMAC().Return("mac", nil)
mockEC2Metadata.EXPECT().VPCID(gomock.Eq("mac")).Return("vpc-id", nil)
mockEC2Metadata.EXPECT().SubnetID(gomock.Eq("mac")).Return("subnet-id", nil)
mockServiceConnectManager := mock_serviceconnect.NewMockManager(ctrl)
mockServiceConnectManager.EXPECT().IsLoaded(gomock.Any()).Return(true, nil).AnyTimes()
mockServiceConnectManager.EXPECT().GetLoadedAppnetVersion().AnyTimes()
mockServiceConnectManager.EXPECT().GetCapabilitiesForAppnetInterfaceVersion("").AnyTimes()
mockServiceConnectManager.EXPECT().SetECSClient(gomock.Any(), gomock.Any()).AnyTimes()

mockDaemonManager := mock_daemonmanager.NewMockDaemonManager(ctrl)
mockDaemonManagers := map[string]dm.DaemonManager{md.EbsCsiDriver: mockDaemonManager}
mockDaemonManager.EXPECT().IsLoaded(gomock.Any()).Return(true, nil).AnyTimes()
mockDaemonManager.EXPECT().LoadImage(gomock.Any(), gomock.Any()).Return(nil, nil).AnyTimes()

dockerClient.EXPECT().SupportedVersions().Return(apiVersions).AnyTimes()

gomock.InOrder(
client.EXPECT().GetHostResources().Return(testHostResource, nil),
mockCredentialsProvider.EXPECT().Retrieve(gomock.Any()).Return(awsv2.Credentials{}, nil),
mockMobyPlugins.EXPECT().Scan().AnyTimes().Return([]string{}, nil),
dockerClient.EXPECT().ListPluginsWithFilters(gomock.Any(), gomock.Any(), gomock.Any(),
gomock.Any()).AnyTimes().Return([]string{}, nil),
client.EXPECT().RegisterContainerInstance(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(),
gomock.Any()).Return("", "", awserr.New("ClientException", "", nil)),
)
mockEC2Metadata.EXPECT().OutpostARN().Return("", nil)

cfg := getTestConfig()
ctx, cancel := context.WithCancel(context.TODO())
// Cancel the context to cancel async routines
defer cancel()
agent := &ecsAgent{
ctx: ctx,
ec2MetadataClient: mockEC2Metadata,
cfg: &cfg,
pauseLoader: mockPauseLoader,
credentialsCache: awsv2.NewCredentialsCache(mockCredentialsProvider),
dockerClient: dockerClient,
mobyPlugins: mockMobyPlugins,
terminationHandler: func(taskEngineState dockerstate.TaskEngineState, dataClient data.Client, taskEngine engine.TaskEngine, cancel context.CancelFunc) {
},
serviceconnectManager: mockServiceConnectManager,
daemonManagers: mockDaemonManagers,
}

exitCode := agent.doStart(eventstream.NewEventStream("events", ctx),
credentialsManager, state, imageManager, client, execCmdMgr)
assert.Equal(t, exitcodes.ExitTerminal, exitCode)
}

func TestRegisterContainerInstanceThrottlingExceptionError(t *testing.T) {
ctrl, credentialsManager, state, imageManager, client,
dockerClient, _, _, execCmdMgr, _ := setup(t)
defer ctrl.Finish()

mockCredentialsProvider := app_mocks.NewMockCredentialsProvider(ctrl)
mockMobyPlugins := mock_mobypkgwrapper.NewMockPlugins(ctrl)
mockEC2Metadata := mock_ec2.NewMockEC2MetadataClient(ctrl)
mockPauseLoader := mock_loader.NewMockLoader(ctrl)

mockPauseLoader.EXPECT().IsLoaded(gomock.Any()).Return(false, nil).AnyTimes()
mockPauseLoader.EXPECT().LoadImage(gomock.Any(), gomock.Any(), gomock.Any()).Return(nil, nil).AnyTimes()
mockEC2Metadata.EXPECT().PrimaryENIMAC().Return("mac", nil)
mockEC2Metadata.EXPECT().VPCID(gomock.Eq("mac")).Return("vpc-id", nil)
mockEC2Metadata.EXPECT().SubnetID(gomock.Eq("mac")).Return("subnet-id", nil)
mockServiceConnectManager := mock_serviceconnect.NewMockManager(ctrl)
mockServiceConnectManager.EXPECT().IsLoaded(gomock.Any()).Return(true, nil).AnyTimes()
mockServiceConnectManager.EXPECT().GetLoadedAppnetVersion().AnyTimes()
mockServiceConnectManager.EXPECT().GetCapabilitiesForAppnetInterfaceVersion("").AnyTimes()
mockServiceConnectManager.EXPECT().SetECSClient(gomock.Any(), gomock.Any()).AnyTimes()

mockDaemonManager := mock_daemonmanager.NewMockDaemonManager(ctrl)
mockDaemonManagers := map[string]dm.DaemonManager{md.EbsCsiDriver: mockDaemonManager}
mockDaemonManager.EXPECT().IsLoaded(gomock.Any()).Return(true, nil).AnyTimes()
mockDaemonManager.EXPECT().LoadImage(gomock.Any(), gomock.Any()).Return(nil, nil).AnyTimes()

dockerClient.EXPECT().SupportedVersions().Return(apiVersions).AnyTimes()

gomock.InOrder(
client.EXPECT().GetHostResources().Return(testHostResource, nil),
mockCredentialsProvider.EXPECT().Retrieve(gomock.Any()).Return(awsv2.Credentials{}, nil),
mockMobyPlugins.EXPECT().Scan().AnyTimes().Return([]string{}, nil),
dockerClient.EXPECT().ListPluginsWithFilters(gomock.Any(), gomock.Any(), gomock.Any(),
gomock.Any()).AnyTimes().Return([]string{}, nil),
client.EXPECT().RegisterContainerInstance(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(),
gomock.Any()).Return("", "", awserr.New("ThrottlingException", "", nil)),
)
mockEC2Metadata.EXPECT().OutpostARN().Return("", nil)

cfg := getTestConfig()
ctx, cancel := context.WithCancel(context.TODO())
// Cancel the context to cancel async routines
defer cancel()
agent := &ecsAgent{
ctx: ctx,
ec2MetadataClient: mockEC2Metadata,
cfg: &cfg,
pauseLoader: mockPauseLoader,
credentialsCache: awsv2.NewCredentialsCache(mockCredentialsProvider),
dockerClient: dockerClient,
mobyPlugins: mockMobyPlugins,
terminationHandler: func(taskEngineState dockerstate.TaskEngineState, dataClient data.Client, taskEngine engine.TaskEngine, cancel context.CancelFunc) {
},
serviceconnectManager: mockServiceConnectManager,
daemonManagers: mockDaemonManagers,
}

exitCode := agent.doStart(eventstream.NewEventStream("events", ctx),
credentialsManager, state, imageManager, client, execCmdMgr)
assert.Equal(t, exitcodes.ExitError, exitCode)
}

func TestMergeTags(t *testing.T) {
ec2Key := "ec2Key"
ec2Value := "ec2Value"
Expand Down
12 changes: 12 additions & 0 deletions agent/app/errors.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,18 @@ func isTransient(err error) bool {
return ok
}

type terminalError struct {
error
}

// isTerminal returns true if the error is already wrapped as an unrecoverable condition
// which will allow agent to exit terminally.
func isTerminal(err error) bool {
// Check if the error is already wrapped as a terminalError
_, terminal := err.(terminalError)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could there be a situation in which this gets wrapped again? Why not use a deeper equality check like errors.Is which inspects the error tree rather than its first layer?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There should not be a case where it gets wrapped again - the terminal wrap should be the last layer as it's handled in this file. So in this case, both errors.Is and the current behaves the same.

Since existing isTransient is implemented similarly - hopefully this optimization shouldn't be a blocker?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is not a blocker, but in the case where we re-use these errors it may be a better long term change to wrap/unwrap these with the errors library.

return terminal
}

// clusterMismatchError represents a mismatch in cluster name between the
// state file and the config object
type clusterMismatchError struct {
Expand Down
Loading