Skip to content

Unified Backups: Investigate Gitaly improvements to Backup

Context

At the time this issue was created, Gitaly has a couple of improvements on the roadmap that can be used for future Backup capability.

There are changes related to how Server-Side backup could work, there are changes related to how the WAL log replication approach could also be used as a base for a PIT recovery, etc.

Scope

We need to consider the following points for the evaluation:

  • A benchmark against existing approach (which is faster, which could be less resource intensive)
  • Which functionality maps well to the Portable backup type, and which maps or works better with the Cloud-based backup type.
  • Whether the WAL logs will be stored as part of the Backup or whether we will "process" the repositories against the WAL logs to get to a certain PIT
  • How does the "observability" of the process look-like (can we provide progress? can we provide meaningful/actionable errors?)
  • How do we validate the repository + WAL logs (processed or stored together) are not corrupted due to the backup process?