Coolify!

[northeastprince]

Migration Plan

Before operation commencement

• Prep and test Litestream restore command
• Spin up temporary Hetzner server that responds 503 with a Retry-After and plaintext response showing a conservative downtime estimate of 5 minutes
• Prep DNS PR changing A record's TTL to a second
• Prep DNS PR changing A record to point to Coolify server

M-minus A record's TTL (currently 10 minutes)

• Merge DNS PR changing TTL

M-minus 0

• Pause all job queues
• Remove app server from LB pool and add temp server
• Run Litestream restore command one last time to sync latest data
• Smoke test on temp endpoint
• Merge DNS PR changing destination to Coolify server
• Unpause queues

[zachlatta]

Thanks so much for doing this. Please use Postgres if you can instead of a volume. The VPS should be considered stateless

[northeastprince]

Won't Postgres be hosted on the VM?

[zachlatta]

Yeah, but it'll be automatically backed up to the s3 backup destination so it can easily be restored.

[northeastprince]

The database is currently synchronously replicated to S3 with Litestream

[zachlatta]

Please don’t have it go to S3, we’re gonna shut down our AWS account. Goal is to have less services and accounts to maintain. Can you have it be automatically backed up to the S3 backups area in Coolify? Thank you!

…

On Jan 14, 2025 at 1:05:42 PM, Matt Almeida ***@***.***> wrote: The database is currently synchronously replicated to S3 with Litestream — Reply to this email directly, view it on GitHub <#611 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHSH6FVQQIOA7KXNU6VBET2KVGXNAVCNFSM6AAAAABVFOL5WKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKOJQG42DSMJUG4> . You are receiving this because your review was requested.Message ID: ***@***.***>

[northeastprince]

Oh, that's awesome! It looks like there might either be multiple Hack Club orgs in Hetzner Cloud (the one I have access to is owned by @maxwofford) or I just can't see the bucket in question. But this app needs persistent (not autocleaned like the backups one) storage anyway to store images and whatnot anyway, so should I just create a separate bucket under the org and use that one?

[northeastprince]

Additionally, I'll need permissions to merge DNS PRs to minimize downtime unless someone else wants to take point on this. Feel free to add anything I missed to the migration plan above!