Repository navigation
Delete handler callback context breaks after multiple invocations #83
Description
It seems that when passing and retrieving data into the callbackContext for the delete handler, after consecutive invocations, it suddenly passes nothing.
I created the following handler:
@resource.handler(Action.DELETE)
def delete_handler(
session: Optional[SessionProxy],
request: ResourceHandlerRequest,
callback_context: MutableMapping[str, Any],
) -> ProgressEvent:
model = request.desiredResourceState
progress: ProgressEvent = ProgressEvent(
status=OperationStatus.IN_PROGRESS,
resourceModel=model,
)
count = 1
if callback_context.get('count'):
count = callback_context.get('count') + 1
LOG.warn("DELETE invocation #{}".format(str(count)))
else:
LOG.warn("First DELETE invocation")
if count < 15:
progress.callbackDelaySeconds = 20
progress.callbackContext = {
'count': count,
}
else:
progress.status = OperationStatus.SUCCESS
return progress
which produces the following logs:
06:40:06
First DELETE invocation
06:40:26
DELETE invocation #2
06:40:46
DELETE invocation #3
06:41:07
DELETE invocation #4
06:41:27
DELETE invocation #5
06:41:47
DELETE invocation #6
06:43:04
Invalid request
Traceback (most recent call last):
File "/var/task/cloudformation_cli_python_lib/resource.py", line 180, in _parse_request
event = HandlerRequest.deserialize(event_data)
File "/var/task/cloudformation_cli_python_lib/utils.py", line 89, in deserialize
event.requestData = RequestData.deserialize(json_data.get("requestData",
{}
))
File "/var/task/cloudformation_cli_python_lib/utils.py", line 54, in deserialize
req_data = RequestData(**json_data)
TypeError: __init__() missing 1 required positional argument: 'systemTags'
The above timing seems to be fairly consistent. I believe the first 3 messages are actually under the same Lambda invocation per local callback rules and that the next 3 messages would be under a second invocation, then the error. I haven't tested with other provider languages so unsure if this is a Go provider issue or Core CLI (service) issue.
Full handler source with example:
cfntypeexample.zip
Activity
iann0036 commentedon Mar 24, 2020
The above also occurs for UPDATE, but not for CREATE
iann0036 commentedon Mar 24, 2020
Unsure if related, but I also can get the following after a long (~5 minute) callback cycle on CREATE:
jaymccon commentedon Mar 24, 2020
I think this is a separate issue, fortunately any errors with publishing metrics are handled without interrupting handler operation