hg.sr.ht is being migrated to a new server soon. Writing up the plans
here for posterity.
1. Build & install the new server (alice5) per normal Alice provisioning
2. Announce planned outage a week in advance
3. Spin up an hg.sr.ht stack and restore from the last backup. This is a
good opportunity to test our backups in action.
4. Test everything!
5. Await planned outage date
6. Set hg.sr.ht¹ to read-only mode (via pgbouncer, probably, and disable
the hg SSH login account)
7. rsync any changes which have occured between steps 3 and 6 to
8. Cut DNS over to hg.sr.ht² and monitor as users get transferred over
9. Monitor hg.sr.ht¹ and shut it off when traffic is more or less done
10. Wait 2 weeks and then decommission hg.sr.ht¹
Things to double check on hg.sr.ht²:
- Is monitoring working? Double check node exporter
- Are backups working?
- Are ZFS snapshots being taken correctly?
- Are ZFS scrubs being run? Double check on the 1st