teamcity: failed test: TestChangefeedTruncateRenameDrop #41566
Assignees
Labels
Milestone
Comments
cockroach-teamcity
added
C-test-failure
|
panic: test timed out after 45m0s |
|
Subsequent build passed, so maybe something is flaky or the timeout was too low. |
|
I |
|
In the stacktrace, I see it the test stuck on a TRUNCATE |
|
Yeah, there's thousands of lines of this in the logs |
|
@aayushshah15 have you tried roachprod-stressrace? It's not frequent on master, but did happen a few times, so I we need to keep pushing (timeouts are really shitty). There's a chance you have to tweak the https://teamcity.cockroachdb.com/project.html?projectId=Cockroach_UnitTests&testNameId=-8759347941440480963&tab=testDetails is the history of that test, fwiw. @ajwerner I think you stressed that test a while ago when it was flaky, any wisdom to share with @aayushshah15? |
|
I would not be stunned if this is fixed by the bug fix that lives in #41842. I'll take this from @aayushshah15. |
This was referenced Oct 25, 2019
|
Here's my best guess. When I did #40581 I forgot to fix https://github.com/cockroachdb/cockroach/blob/master/pkg/sql/truncate.go#L193. This is problematic because it means that This isn't a changefeed bug per se. |
craig bot
pushed a commit
that referenced
this issue
Nov 25, 2019
42650: sql: stop observing the CommitTimestamp in TRUNCATE r=ajwerner a=ajwerner In #40581 we stopped observing the commit timestamp to write it into table descriptors. In this change I overlooked (rather forgot) about this additional place in the code where we observed the commit timestamp. As far as I can tell we don't read this field anywhere ever. Furthermore we know that the the table descriptor in question to which we are referring must be alive and equal to the provided value at the timestamp at which it was read due to serializability. In short, this minor change continues to populate the field with a sensible value and will permit TRUNCATE to be pushed. Fixes #41566. Release note (bug fix): Long running transactions which attempt to TRUNCATE can now be pushed and will commit in cases where they previously could fail or retry forever. 42746: roachtest/cdc: fix cdc/bank and cdc/schemareg r=nvanbenschoten a=nvanbenschoten Fixes #41177. Fixes #42690. These were both broken by #41793 because prior versions of crdb didn't support the `WITH diff` option. Co-authored-by: Andrew Werner <[email protected]> Co-authored-by: Nathan VanBenschoten <[email protected]>
ajwerner
added a commit
to ajwerner/cockroach
that referenced
this issue
Nov 26, 2019
In cockroachdb#40581 we stopped observing the commit timestamp to write it into table descriptors. In this change I overlooked (rather forgot) about this additional place in the code where we observed the commit timestamp. As far as I can tell we don't read this field anywhere ever. Furthermore we know that the the table descriptor in question to which we are referring must be alive and equal to the provided value at the timestamp at which it was read due to serializability. In short, this minor change continues to populate the field with a sensible value and will permit TRUNCATE to be pushed. Fixes cockroachdb#41566. Release note (bug fix): Long running transactions which attempt to TRUNCATE can now be pushed and will commit in cases where they previously could fail or retry forever.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The following tests appear to have failed on release-19.2 (testrace): TestChangefeedTruncateRenameDrop, TestChangefeedTruncateRenameDrop/sinkless, TestChangefeedTruncateRenameDrop/enterprise
You may want to check for open issues.
#1537922:
Please assign, take a look and update the issue accordingly.
The text was updated successfully, but these errors were encountered: