Attendees:
Squeaky Pancakes notplants Calix
planning the migration
We can now pay for Kim Sufi hosting month by month (since April 5th 2022 we are doing this)
for now we can continue doing this until we figure out where we really want to move next (without the extra work of an intermediary migration)
postgres dump is taking forever
squeaky pancakes has been working on an upgrade. for some reason, running postgres_dump was taking a very long time (it seems that its actually hanging, file hasn't changed size in a while)
we all watch postgres hanging in terminal together, there is a feeling of mystery and anticipation
restoring backups
we briefly discuss whether we could test restoring a backup sometime, in part because we are not sure why the most recent backup squeaky has done had a different size (30gb instead of 10gb) from the nightly backups we run (we want to confirm that the nightly backup actually has everything it needs). we speculate that possibly the two backups are just different formats (due to the nightly backup passing -F flag to pg_dump)
we decide we can just test restoring the backup when we eventually migrate the server, since we will need to do a backup and restore then
which server to migrate to
calix says that it still could be a couple months until servers.coop is ready. they also share that fediverse has created a list of hosting providers which are cheaper than servers.coop (which they will also share with sunbeam, so we can in full transparency make the decision that feels best to the group)
notplants shares that they would still prefer to use servers.coop, but are happy to see the list and all possible alternatives as well
can we lose 7 hours of data?
we have a successful backup from 7 hours ago. we discuss whether its ok to use this one, or if we need to ensure we have a new backup right now, which includes any posts any users may or may have not made in the last 7 hours.
restore is working!
successfully using pg_restore to restore pg_dump
we are impressed that postgres is able to do this across postgres versions (https://www.theregister.com/2021/12/06/mysql_a_pretty_poor_database/)
we just had to create a new database with name masotodon_production before running the pg_restore command
TODO / next steps:
- get the server running and make sure it works (ty squeaky and calix)
- next time we need to decide where we will migrate to (hopefully getting the input of others) … noptlants will make a post about our interim proposal to go with kim sufi month to month for now
- what are the services that sunbeam is hosting, and are their some to retire, and some to add
Also see: