Consistent Backups: Discovery on how to handle PostgreSQL backup using WAL log shipping solution
Context
The approach to backup larger installations will not rely on the same approach we today of triggering a snapshot and using that to export the data via pg_dump. It will most likely rely on WAL log and some solution that allows us to navigate and restore data at a certain point-in-time.
Discovery
In that type of scenario, we need to figure out how to handle the WAL log solution when creating a final archive for storage:
- We need to figure out if we can preprocess the data to be ready at the point-in-time we selected
- In case we can't preprocess in advance, we need to figure out how to process it at time of restoring (and how to persist the information about the correct point-in-time inside the backup)
- Do we need to use the same version for PostgreSQL WAL log backup tool when restoring?
- Do we need to restore the data to the exact same PostgreSQL version?
- In case the exact same PostgreSQL version is required, can we preprocess the data when creating the backup to make it portable to current + newer versions (like the
pg_dump
solution allows us today?)
Edited by Gabriel Mazetto