statusccl: fix flaky TestTenantStatusAPI txn_id_resolution #100817
Conversation
|
This change is |
There was a problem hiding this comment.
Reviewed all commit messages.
Reviewable status:complete! 0 of 0 LGTMs obtained (waiting on @j82w)
pkg/ccl/serverccl/statusccl/tenant_status_test.go line 1231 at r1 (raw file):
} return nil }, 1*time.Minute)
Is there a concern this wait time could increase too much the time of the test and it could cause issue under stress? Should this test be skipped under stress?
|
Previously, maryliag (Marylia Gutierrez) wrote…
It's already skipped. |
|
Previously, j82w (Jake) wrote…
The extended time is probably not necessary now that it actually retries on default fingerprint, but I made the change to be on the safe side. |
The txn resolution add a txn id to the cache, and then in a separate call updates the txn id with the fingerprint. The test previously assumed if there was a record it would always have a valid fingerprint, but that is not true. The test is modified to add retry logic to wait for the fingerprint to be updated. The retry time was set to always be 1 minute, because SucceedsSoon only waits 5 seconds for race scenario which is not enough time for the resolution to complete. Epic: none part of: #99770 Release note: none
There was a problem hiding this comment.
Reviewed 1 of 1 files at r1.
Reviewable status:
|
bors r+ |
|
Build failed: |
|
bors r+ |
|
Build succeeded: |
The txn resolution add a txn id to the cache, and
then in a separate call updates the txn id with the
fingerprint. The test previously assumed if there
was a record it would always have a valid fingerprint,
but that is not true. The test is modified to add retry
logic to wait for the fingerprint to be updated. The
retry time was set to always be 1 minute, because
SucceedsSoon only waits 5 seconds for race scenario
which is not enough time for the resolution to complete.
Epic: none
part of: #99770
Release note: none
Epic: CRDB-25476