Should I click? I fear breaking everything.
Go ahead. Don’t worry. I have done upgrades in the past. It doesn’t take to long.
Does it work like the common WordPress setup, in that the webserver user has write access to the application files?
This site is back online.
I’ve also made myself an administrator, so I can check out the backups etc.
Yep! I reckon. It’s actually quite cool learning more about how Discourse is deployed, because it’s probably a setup that’s just going to become more common (i.e. only requiring Docker, and nothing else).
I’ll try to document what I can on wikitech. And we need to sort out backups and exports.
Thank you so much @samwilson!
For the record in this thread, this is what happened when I tried to upgrade:
Talking of Phab, we should probably create a project over there for tracking this stuff. If there’s going to be more of it.
Thanks for all the efforts. The discourse installation is a test
installation not a production installation. We now know for sure that prior
to upgrading a production installation an upgrade test on a test
installation is necessary, because of a growing list of dependencies.
Nonetheless, Quim can enjoy discourse again
Probably, but I think it is better to wait at least until the beginning of the discourse-mediawiki pilot.
If / when we agree on moving from a pilot in wmflabs.org to production, a workflow for software upgrades needs to be carefully planned, yes. In the meantime, we will need to be careful and read the release notes.
The main problem with the upgrade that broke is that it was a too big jump, from v1.5 to v1.9. Similarly to Phabricator, Discourse releases updates every few weeks. Similarly to Wordpress, most of the times upgrading is possible by clicking a button in the administrators’ dashboard. But from time to time the required dependencies need an upgrade too. These changes in backend requirements are announced in the release notes, and I even think they are supposed to appear in the dashboard as blockers. But as said, this time the ump of version was probably too big for a simple button click.
The jump was too big. It happened because of low traffic on this test site, and I hadn’t checked in for a long time. Now there are three admins. So we will see when the next upgrade breaks