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

feat(snapshot-backfill): measure accumulated row count for each vnode stream #20016

Merged
merged 1 commit into from
Jan 3, 2025

Conversation

wenym1
Copy link
Contributor

@wenym1 wenym1 commented Jan 3, 2025

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

Part of #19720

Measure the row count of each vnode stream, so that the row count can be saved in the state to be introduced in #19720.

This PR will also fix a bug found in the CI #19720. In short, the bug is caused by the use of FuturesUnordered. In FuturesUnordered, pending futures will be stored separately and will not be polled until a special cx generated in FuturesUnordered::poll_next is awakened. However, when we use Peekable::peek to peek the latest progress of the stream, the special cx will be overwritten when we poll the Peek future, and then even the future is ready after peeked, it is still not in the ready queue, and will not be polled and get stuck forever.

Tests are added to cover the bug.

Checklist

  • I have written necessary rustdoc comments.
  • I have added necessary unit tests and integration tests.
  • I have added test labels as necessary.
  • I have added fuzzing tests or opened an issue to track them.
  • My PR contains breaking changes.
  • My PR changes performance-critical code, so I will run (micro) benchmarks and present the results.
  • My PR contains critical fixes that are necessary to be merged into the latest release.

Documentation

  • My PR needs documentation updates.
Release note

Copy link
Contributor Author

wenym1 commented Jan 3, 2025

Copy link
Collaborator

@hzxa21 hzxa21 left a comment

Choose a reason for hiding this comment

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

LGTM

// The pending futures will be stored in a separate linked list and will not be polled until the special cx is awakened
// and move the awakened future from the linked list to a ready queue.
//
// However, here if we use `FuturesUnordered::iter_mut` to in place access the pending futures and call `peek` directly,
Copy link
Collaborator

Choose a reason for hiding this comment

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

We use iter_mut and peek prior to this pr as well. Does it mean that this issue existed before but we didn't hit it in test?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes, peek is used in for_vnode_pk_progress. The method is implemented but not called yet.

@wenym1 wenym1 added this pull request to the merge queue Jan 3, 2025
Merged via the queue into main with commit e08b0f9 Jan 3, 2025
35 of 37 checks passed
@wenym1 wenym1 deleted the yiming/vnode-stream-row-count branch January 3, 2025 10:42
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.

2 participants