Open ccampbell-crl opened 1 year ago
cc @cockroachdb/disaster-recovery
This is already possible, though not obvious.
A full backup has syntax BACKUP <target> INTO <storage>
An inc backup has syntax BACKUP <target> INTO <full_backup> IN <storage>
full_backup
can either be LATEST
, or an explicit directory in the storage bucket that contains a full backup. The key differentiator here is IN
.
That said this is an extremely subtle distinguishing feature. We could consider ways to make this more observable.
@benbardin Isn't the IN
clause always in the description once the job has run?
When you look at the jobs after it has run, I thought I remembered that the description of the full is updated to reflect where it actually backed up to, i.e. it is shown as BACKUP INTO date-based-path IN collection
, which is identical to if it had been an incremental to that location.
Also incremental will not show up as LATEST
after they have been run: the job description is updated with the resolved path once it is resolved.
I believe there is a subtle way you can tell, which is that the AOST time will match the date-based-path time on the full, but not for subsequent later incremental into that same path, but that's even more subtle than the IN
clause.
When you look at the jobs after it has run, I thought I remembered that the description of the full is updated to reflect where it actually backed up to, i.e. it is shown as BACKUP INTO date-based-path IN collection, which is identical to if it had been an incremental to that location.
TIL!
I believe there is a subtle way you can tell, which is that the AOST time will match the date-based-path time on the full, but not for subsequent later incremental into that same path, but that's even more subtle than the IN clause.
🤣
TSE has noticed difficulty in easily identifying if a backup in our crdb_internal.jobs.txt is a full backup or an incremental backup. Customers also may have difficulty identifying backup type by looking in the DBConsole.
The request is for engineering / Disaster Recovery to come up with a way to easily identify full backups vs incremental backups in both a debug.zip and the DBConsole.
One example would be to have two different “types” of backups in the crdb_internal.jobs.txt file in a debug, “Backup - Full” and “Backup - Incremental”
Jira issue: CRDB-23679