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.

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

Bugfixes from omnibus 2 PR #3034 for connector sink #3042

Merged
merged 5 commits into from
Nov 11, 2019

Conversation

slfritchie
Copy link
Contributor

From omnibus 2 PR #3034.

Merge after PR #3039.

@slfritchie
Copy link
Contributor Author

Failures of the integration-tests-testing-correctness-tests-aloc_sink test are due to an interdependence on PR #3039.

@slfritchie
Copy link
Contributor Author

Fixes #3023 (when combined with PRs #3039 and #3041)

@slfritchie
Copy link
Contributor Author

I've rebased this PR onto today's master branch, which should make the single CI failure go away.

@slfritchie
Copy link
Contributor Author

Chat with John: Scott will update the base of this PR after merging several outstanding PRs. Some/all autoscale changes made here may no longer be necessary.

@slfritchie
Copy link
Contributor Author

@jtfmumm: This PR looks like it plays well with today's master branch @ commit 4323b71; the bugfixes for connector sink 2PC, avoiding Pony runtime backpressure, and the new "workers left" message to the connector sink are all important.

* Add debug diagnostics
* Fix sink phase errors at early startup/InitSinkPhase
* Fix sink phase errors with autoscale barrier processing
* Add 2PC WorkersLeft message sending
* Add  _phase.early_prepare_for_rollback()
* be rollback(): Send 2PC commands only if connected & 2PC intro done
* Remove use of Pony runtime backpressure by sink when throttled
* Add WorkerLeftMsg unit tests
Fixes #3015

Verified by using `./master-crasher.sh 1 crash-sink`, let the sink crash
and restart a couple of times, then verify in /tmp/sender.out that there
are new "acked Stream" messages where "is_open=True" and the
"point_of_ref" value continues to increase.
* Move check for calling checkpoint_complete()

* Avoid matching directly on the sink phase.

* Add new class EarlySinkPhase to work around a chicken-and-the-egg
  problem caused by needing a valid Sink phase and not yet having
  a fully-initialized `Sink` object to pass to the constructor of
  the phase.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants