Restart from #492
Merged
Restart from #492
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
After a failure the KCL would instead restart from what the ShardConsumer says it last processed.
Extracted the InternalSubscriber to ShardConsumerSubscriber to make testing easier. Added tests for the ShardConsumerSubscriber that verifies error handling and other components of the class. Added tests that verify the restart from behavior.
When the blocking queue is full it would normally enter into a fully parked state, but would continue to hold the lock. This changes the process to only block for a second when attempting to enqueue a response, and if it doesn't succeed check to see if it's been reset before attempting again.
Changed the locking around the restart to use a reader/writer lock instead of single lock with a yield. Changed how the fetcher is reset to not restart from an advanceIteratorTo which would retrieve a new shard iterator. Instead the resetIterator method takes both the iterator to start from, the last accepted sequence number, and the initial position.
Changed the test to wait for the queue to reach capacity before restarting the PrefetchRecordsPublisher. This should mostly ensure that calling restartFrom will trigger a throw of a PositionResetException. Added @VisibleFortest on the queue since it was already being used in testing.
In the test the TestPublisher is accessed from two threads: the test thread, and the dispatch thread. Both have the possibility of calling send() under certain conditions. This changes it so that only one of the threads can actively be sending data at a time. TestPublisher#requested was changed to volatile to ensure that calling cancel can correctly set it to zero.
This test is somewhat of an odd case as it intends to test what happens when nothing is dispatched to the ShardConsumerSubcriber for some amount of time, but data is queued for dispatch. To do this we block the single thread of the executor with a lock to ensure that items pile up in the queue so that should the restart work incorrectly we will see lost data.
pfifer
added a commit
to pfifer/amazon-kinesis-client
that referenced
this pull request
Feb 18, 2019
https://github.com/awslabs/amazon-kinesis-client/milestone/29 * Fixed handling of the progress detection in the `ShardConsumer` to restart from the last accepted record, instead of the last queued record. * awslabs#492 * Fixed handling of exceptions when using polling so that it will no longer treat `SdkException`s as an unexpected exception. * awslabs#497 * awslabs#502 * Fixed a case where lease loss would block the `Scheduler` while waiting for a record processor's `processRecords` method to complete. * awslabs#501
sahilpalvia
pushed a commit
that referenced
this pull request
Feb 18, 2019
https://github.com/awslabs/amazon-kinesis-client/milestone/29 * Fixed handling of the progress detection in the `ShardConsumer` to restart from the last accepted record, instead of the last queued record. * #492 * Fixed handling of exceptions when using polling so that it will no longer treat `SdkException`s as an unexpected exception. * #497 * #502 * Fixed a case where lease loss would block the `Scheduler` while waiting for a record processor's `processRecords` method to complete. * #501
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Issue #, if available:
Description of changes:
If something goes wrong between the
ShardConsumer, and theRecordsPublisherit's possible that theShardConsumermay cancel the connection to theRecordsPublisher.RecordsPublisher's are stateful so it's possible the state ofRecordPublisherhas advanced further through the shard than theShardConsumerhas actually processed. This could cause theShardConsumerto skip records.The
ShardConsumerwill now be able to ask theRecordsPublisherto restart from a specific location, instead of restarting from the last retrieved batch of records.By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.