Repository navigation
[profiler] Remove support for legacy profiles produced pre 0.244 (Feb 2026) - #7756
Conversation
|
What I ran
Gates
Besides these, the removal and the migration of the tests to the new metric names look correct. |
… 26) Signed-off-by: Mihai Budiu <mbudiu@feldera.com>
|
Verdict: not this PR's fault (merge-queue drop was collateral). High confidence — the diff here only touches The only real failure was in
This test shares a fixed cache prefix across CI runs for its checkpoint fixture, and has a history of cross-run races on that shared state (#7136, #7572). The error here comes from a storage-corruption guard that fires exactly on a torn/partial state of that shared cache — consistent with another run touching the same prefix concurrently, not with anything this PR changed. Filed as #7763 with the evidence and a race hypothesis. Next step: re-queueing should pass; this PR's own changes aren't implicated. |
|
Verdict: not this PR's fault. High confidence — the diff here only touches Failing job: Adapter Tests / Test Adapters (arm64) This is a known pre-existing bug introduced on Next step: re-queueing should pass; this is unrelated to this PR's changes. |
Verdict: not this PR's fault (pre-existing flake on
|
| Failed job | Adapter Tests / Test Adapters (amd64) |
| Failing test | transport::kafka::ft::output::tests::real::restart_skips_a_commit_whose_reply_was_lost |
| Panic | log line 1676: the consumer must reach the high watermark after the broker's own transaction.timeout.ms (60s) expired while the test's proxy was holding the commit reply |
This is the same new Kafka FT output test feature (commit 6ad915e99, merged 2026-10-28) already flaking other merge-queue runs under issue #7760 — just a different test/symptom there (spool-not-released assertion vs. this timeout). Filed as a separate issue since the test and failure mode differ: #7765.
Next step: safe to re-queue. This looks like a timing-budget race in the test itself (proxy hold duration vs. the broker's 60s transaction.timeout.ms), unrelated to your changes.
Karakatiza666
left a comment
There was a problem hiding this comment.
LGTM. I am happy we can do this
|
Same failure already explained in a previous comment on this PR: the
PR's diff still only touches |
Linked Issue
Fixes #7198
That code is hard to read and maintain.