Backup and restore from remote sites?

[whoschek]

Interesting project, congrats!

Does it make sense to use this in combination with bzfs and/or bzfs_jobrunner (of which I'm the author), and if so what would be the best way to do so?

I'd assume you'd probably omit any bzfs functionality related to taking snapshots as velo already has it's own semantics wrt. creating and deleting snapshots. But maybe as a means for backup and recovery of velo dbs to/from remote sites?

Thoughts?

[elitan]

Hey @whoschek! Thanks for opening this issue and for the kind words about the project. I took a look at bzfs and it's really impressive - the incremental replication and zero-dependency approach is exactly the kind of tool I appreciate.

I've been thinking about backup/restore for Velo and wanted to share my current thinking and get your feedback.

Current State

Right now Velo has comprehensive local backup capabilities:

• Application-consistent snapshots (via CHECKPOINT)
• WAL archiving for PITR
• Local point-in-time recovery

What's missing is remote backup for disaster recovery.

Approaches I've Been Considering

I see two main paths:

1. S3-compatible storage (simple path)

• Works with AWS S3, Cloudflare R2, Backblaze B2, MinIO, etc.
• Dead simple setup (just bucket + credentials)
• Very cheap ($0.50-2/month for typical databases)
• Zero maintenance (serverless)
• Trade-off: Full backups only (no efficient incrementals), slower for large databases

2. Remote ZFS replication (power user path)

• This is where bzfs integration would shine
• Efficient incremental replication
• Fast restores
• Perfect for large databases (>100GB)
• Trade-off: Requires managing a remote ZFS server, more complex setup

One Idea I Had

Start with S3 support to cover the simple/common cases (small-medium databases, daily backups, users who want zero hassle), then add bzfs integration for power users who need efficient incrementals and have large databases.

But I'm really open to other approaches!

Questions for You

1. What's your primary use case? Large production databases that change frequently?
2. Would you find value in S3 backup even if bzfs integration came later, or is bzfs the main thing you're interested in?
3. Are you thinking of this primarily as:
   • Disaster recovery (backup main branches to remote)?
   • Replication (sync production to staging servers for branching)?
   • Both?

I'd love to collaborate on this. If bzfs integration is the priority, I'm open to that approach - just want to understand the use cases better to design the right solution.

What do you think?