-
Notifications
You must be signed in to change notification settings - Fork 3.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
ccl/multiregionccl: TestMrSystemDatabase failed #122790
Comments
We previously, this was being hit because the transition to a multi-region database was not well tolerated in the system database. The latest variant is something else, since I don't see any messages from the counting logic |
This is also limited to the system database case, which isn't fully supported as of yet, so dropping the release-blocker tag. Still trying to reproduce it since, its not clear why the SELECT is seeing this error. |
ccl/multiregionccl.TestMrSystemDatabase failed on master @ 2f498d0abcf4130e5c872b73200086d6a3af9f02:
Parameters:
|
This is happening because we accidentally reverted: https://github.com/cockroachdb/cockroach/pull/117139/files. Let me get a PR to address it again |
Previously, when converting the system database to multiregion its possible for table statistics to contain the existing type of crdb_region as bytes. This could happen if automatic statistics collection happened concurrently with the conversion to a multi-region system database. The conversion had logic to clear table statistics, but it was still possible for statistics collection to happen in between. This could cause queries against RBR system tables to fail because, since the table_statistics type information no longer matches with the table descriptor after. We started seeing this for the system database inside TestMrSystemDatabase, once conversion was added for the system tenant. To address this, this patch first adds extra logic in the schema changer to force a refresh of stats on system tables, which will force a refresh of statistics after the schema change, in case a stats refresh occurs before the job completes. We also modify the TestMrSystemDatabase to intentionally generate stats before changing the system database under the system tenant to avoid the risk of hitting this issue. With these changes we expect the test to no longer flake and any real world occurrence to be less transient. Fixes: cockroachdb#122790 Release note: None
123905: sql: refresh stats for multi-tenant system database conversions r=fqazi a=fqazi Previously, when converting the system database to multiregion its possible for table statistics to contain the existing type of crdb_region as bytes. This could happen if automatic statistics collection happened concurrently with the conversion to a multi-region system database. The conversion had logic to clear table statistics, but it was still possible for statistics collection to happen in between. This could cause queries against RBR system tables to fail because, since the table_statistics type information no longer matches with the table descriptor after. We started seeing this for the system database inside TestMrSystemDatabase, once conversion was added for the system tenant. To address this, this patch first adds extra logic in the schema changer to force a refresh of stats on system tables, which will force a refresh of statistics after the schema change, in case a stats refresh occurs before the job completes. We also modify the TestMrSystemDatabase to intentionally generate stats before changing the system database under the system tenant to avoid the risk of hitting this issue. With these changes we expect the test to no longer flake and any real world occurrence to be less transient. Fixes: #122790 Release note: None Co-authored-by: Faizan Qazi <faizan@cockroachlabs.com>
ccl/multiregionccl.TestMrSystemDatabase failed on master @ 347cdc76d4c5abb2e872f325e944337a46b5883f:
Parameters:
attempt=1
race=true
run=3
shard=3
Help
See also: How To Investigate a Go Test Failure (internal)
This test on roachdash | Improve this report!
Jira issue: CRDB-38068
The text was updated successfully, but these errors were encountered: