bookkeeper icon indicating copy to clipboard operation
bookkeeper copied to clipboard

[improve] Force Reschedule Auditor tasks

Open wenbingshen opened this issue 2 years ago • 5 comments

Motivation

This PR follows up the comments in this PR #3945

Now when we manually trigger auditor check, we need to restart Auditor or trigger a round of auditor election to take effect.

So I hope we can trigger auditor check more smoothly. Support for Auditor Tasks rescheduling is available.

Changes

By registering the specified zk node, the callback logic of Auditor tasks rescheduling is triggered.

This PR adds the following changes:

  1. Add specified zk node for reschedule auditor tasks.
  2. The callback logic for reschedule auditor tasks.
  3. Add metric for reschedule auditor tasks.
  4. Add force flag to ForceAuditorChecksCmd to emit reschedule auditor tasks and cmd unit tests.
  5. Add Auditor tasks to be rescheduled unit tests.

BP

BP #4025 email: https://lists.apache.org/thread/w1rqy6t2jr3qrrzxjh14rjllpgw97v98

wenbingshen avatar May 20 '23 06:05 wenbingshen

@hangc0276 @horizonzy @zymap @StevenLuMT This PR is ready for review. I have improved the remaining functions & add metric & add trigger cmd and unit tests & add Auditor unit tests. PTAL. Thanks.

wenbingshen avatar May 24 '23 14:05 wenbingshen

rerun failure checks

wenbingshen avatar Jun 13 '23 03:06 wenbingshen

@hangc0276 Can we include this future in 4.16.2?

wenbingshen avatar Jun 15 '23 05:06 wenbingshen

The function is still quite complicated,or add a BP to discuss?

@StevenLuMT Thanks. I have started a BP to discuss it. BP #4025 email: https://lists.apache.org/thread/w1rqy6t2jr3qrrzxjh14rjllpgw97v98

wenbingshen avatar Jul 10 '23 06:07 wenbingshen