Fix consumer dead state bug #364
Conversation
| runActionWithExpectedResult(actionName, "dat/multipleValueTypes.json", expectedOutput, false) | ||
| } | ||
|
|
||
| it should "create a trigger, delete that trigger, and quickly create it again with successful trigger fires" in withAssetCleaner(wskprops) { |
There was a problem hiding this comment.
Test fails when used against provider without the included updates.
| logging.info('[{}] Found a change to an existing trigger'.format(change["id"])) | ||
|
|
||
| if existingConsumer.desiredState() == Consumer.State.Disabled and self.__isTriggerDocActive(document): | ||
| if existingConsumer.desiredState() == Consumer.State.Dead and self.__isTriggerDocActive(document): |
There was a problem hiding this comment.
For background info, there is a nanny running periodically that deletes consumers that are in a dead state. Since the consumer is already in a dead state waiting for the nanny to close the process, a quick recreate of the consumer -- within a 2 second time period -- will be ignored. Instead of ignoring the recreate we will manually delete the consumer so that it can be recreated. Normally a graceful shutdown using the nanny is preferred, but we have no other choice here.
| # Record the sequence in case the changes feed needs to be | ||
| # restarted. This way the new feed can pick up right where | ||
| # the old one left off. | ||
| self.lastSequence = change['seq'] |
There was a problem hiding this comment.
Moved this in case of an unexpected exceptions that could happen, allowing us to pick up where we left off.
2 participants
When an existing consumer is deleted and then quickly recreated, that consumer will get stuck in a dead state. This can happen when using wskdeploy with a manifest that undeploys triggers and then immediately recreates those triggers quickly.