Hi David,

I've submitted a PR for FLINK-38534 (flaky LocalRecoveryTest): 
https://github.com/apache/flink/pull/27767

The test was intermittently timing out in CI. Based on my investigation, the 
issue appears to be that checkpoints were triggered before tasks fully 
transitioned to RUNNING state. The fix
adds synchronization to ensure state transitions complete before checkpoint 
triggers, following a similar approach to FLINK-39182.

Verified locally with 300+ iterations. Ready for review.

Best,
Mukul

-----Original Message-----
From: David Anderson <[email protected]>
Sent: 11 March 2026 21:16
To: dev <[email protected]>
Subject: [SUMMARY] Flink 2.3 release sync -- 3 weeks to feature freeze

We had another Flink 2.3 release sync on March 10th / March 11th (depending on 
your time zone).

The important points are:

* The feature freeze is coming in 3 weeks, at the end of March. A couple of 
FLIPs are in danger of not being completed.
* The Azure CI builds are frequently failing.

If you've been assigned to fix flaky tests, please follow through. These 
tickets need attention:

FLINK-39145 assigned
FLINK-38727 assigned
FLINK-39117 assigned
FLINK-38970 assigned
FLINK-38872 PR in review
FLINK-38534
FLINK-38507

[1] https://cwiki.apache.org/confluence/display/FLINK/2.3+Release

Reply via email to