backupccl: empty full followed by incremental backup can result in use of the wrong codec #115773
Labels
A-disaster-recovery
C-bug
Code not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior.
T-disaster-recovery
Projects
An empty full backup captures no Spans in the backup manifest. If this full backup is then followed by an incremental that does capture some data, the codec we use in the key rewriter is incorrect. This is because we resolve the codec based on the Spans in the full backup manifest, of which there are none, causing us to default to a SystemCodec. Using a system codec when running inside an application tenant generates an error when decoding the key and causes the restore to fail. We should be inferring the codec from the first non-empty backup manifest.
Jira issue: CRDB-34204
The text was updated successfully, but these errors were encountered: