Skip to content

Conversation

@stevendanna
Copy link
Collaborator

@stevendanna stevendanna commented Oct 20, 2025

When using tenants, this test appears more prone to flakes. The flakes seem to come from two sources:

When using tenants, this test appears more prone to flakes. The flakes
seem to come from two sources:

  1. A lease would move off of node 1 and thus we'd see a
    "local read" succeed. Here we try to make this less likely by turning
    off the lease, replicate, and split queues and adding an early
    assertion that all of the leaseholders live on n1.

  2. The schema change the test performs would take > 10s, which then
    invalidates many later assumptions that 10s in the past should be
    strictly before the schema change. We've kicked the can here and
    bumped it to 20s.

Unfortunately, this still doesn't seem to be enough as we still
occasionally see this fail. One note is that this test is intended to be
skipped under stress, but in CI it is still run under stress because of
changes to the meaning of skip.UnderStress.

Informs #154710
Release note: None

@cockroach-teamcity
Copy link
Member

This change is Reviewable

@stevendanna stevendanna force-pushed the ssd/deflake-TestBoundedStalenessDataDriven branch from 138ee5e to 6b3786d Compare October 20, 2025 11:07
@stevendanna
Copy link
Collaborator Author

Maybe this will help us get through extended CI: #154939

When using tenants, this test appears more prone to flakes. The flakes
seem to come from two sources:

1. A lease would move off of node 1 and thus we'd see a
   "local read" succeed. Here we try to make this less likely by turning
   off the lease, replicate, and split queues and adding an early
   assertion that all of the leaseholders live on n1.

2. The schema change the test performs would take > 10s, which then
   invalidates many later assumptions that 10s in the past should be
   strictly before the schema change. We've kicked the can here and
   bumped it to 20s.

Unfortunately, this still doesn't seem to be enough as we still
occasionally see this fail. One note is that this test is intended to be
skipped under stress, but in CI it is still run under stress because of
changes to the meaning of skip.UnderStress.

Informs cockroachdb#154710
Release note: None
@stevendanna stevendanna force-pushed the ssd/deflake-TestBoundedStalenessDataDriven branch from 6b3786d to 228c97d Compare October 27, 2025 15:58
Copy link
Collaborator

@arulajmani arulajmani left a comment

Choose a reason for hiding this comment

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

:lgtm:

@arulajmani reviewed 2 of 2 files at r1, all commit messages.
Reviewable status: :shipit: complete! 1 of 0 LGTMs obtained (waiting on @stevendanna)

@stevendanna
Copy link
Collaborator Author

bors r=arulajmani

@craig
Copy link
Contributor

craig bot commented Oct 28, 2025

@craig craig bot merged commit 707486e into cockroachdb:master Oct 28, 2025
33 of 34 checks passed
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.

3 participants