The storage engine often uses cluster versions for migrations and to impose restrictions on clients. There's new subtlety now that under unified architecture a tenant's cluster version may lag behind the system cluster version. We should add a comment to the places where we determine the appropriate table format to use when building an sstable. Maybe there's additional work we should consider.
The storage engine often uses cluster versions for migrations and to impose restrictions on clients. There's new subtlety now that under unified architecture a tenant's cluster version may lag behind the system cluster version. We should add a comment to the places where we determine the appropriate table format to use when building an sstable. Maybe there's additional work we should consider.
Jira issue: CRDB-33932