Publication date: September 12, 2023
Description
A CockroachDB cluster using row-level TTL for a table schedules background jobs that periodically delete expired data and add a field to the table metadata to indicate that the row-level TTL feature has been enabled for the table. Upgrading the cluster from CockroachDB v22.2.x to v23.1.9 erroneously deletes the field from the table metadata, causing the scheduled background jobs for that table to crash upon upgrade.
This issue affects clusters that meet all of the following conditions:
- The cluster was running CockroachDB v22.2.x.
- The cluster is upgraded to v23.1.9.
- The cluster contains any databases using the row-level TTL feature for their tables.
Statement
This is resolved in CockroachDB by #110364, which prevents deletion of the TTL field upon upgrade.
The fix will be available in CockroachDB v23.1.10.
This public issue is tracked by #110363.
Mitigation
Users of CockroachDB v22.2.x using row-level TTL should not upgrade to 23.1.9 and are advised to either upgrade to 23.1.8 or wait for 23.1.10.
If you are not using row-level TTL, or if you are using row-level TTL but are already on version 23.1.x, this recommendation does not apply.
If you think you have already encountered this problem, please reach out to the support team for assistance.
Impact
Customers using row-level TTL for their cluster and upgrading from v22.2.x to 23.1.9 can experience nodes crashing due to background jobs for TTL failing. Customers who are on v22.2.x with TTL enabled should not upgrade to 23.1.9 and are advised to either upgrade to 23.1.8 or wait for 23.1.10.
Questions about any technical alert can be directed to our support team.
Was this helpful?
