After updating via the gui this morning, the site is down. Seems like it tried to update from Postgres V15 to V18, but failed. So I rebuilt from within the app via launcher, no help. Also a complaint about VIPS /usr/local/lib/libvips.so.42
Any ideas or is it just me?
It’s always worth trying a second rebuild, especially with a database version bump. See the nearby topic
If you still have trouble, please paste 100 lines or so of the output that you saw, preferably in a code block.
As Chris notes, although the advice might be a little late in the day for you:
I did try that but of course did not see the thread you mention. I had made a backup before I started the update. So I was able just now to get it working by manually updating to PG V18 within the container, and also installing and pointing to a later version of vips in the container:
The site is working now, I suppose.. Then I ran a full rebuild and it succeeded, so hopefully all is well now.
vips-8.18.4
PostgreSQL 18.4
However, I will go in and read the thread you posted in detail, as I have another instance on another machine to update also.
THANK YOU Ed_S!
interesting that I already have in the app.yml file from Oct 2025:
templates:
Am I reading something incorrectly from the other topic?
I think that means you have the default choice, which is not to pin postgres to a particular version, but take whatever Discourse likes.
(People will pin a particular version when they see the new version coming and want not to take it. But they need to act before it arrives unexpectedly.)
OK, sigh, so I am not sure then what went wrong with my update this morning, why it failed. I am leery of upgrading the other site until I understand why this site failed, as I do not want to repeat the failure, especially since I am not fond of manually upgrading things within the container as I did to get this first site running.
My guess is that this update, from today, which introduces the version 18 of postgres, will need two rebuilds. From the web interface it’s only possible to do one rebuild, and if I’m right about that, a web interface update will always appear to fail. In fact, it didn’t fail, but only went half way.
This guess might be wrong - I can’t easily test it.
More than one person, including me, has said they no longer use the web update mechanism, because if it does fail then you need to visit the command line anyway. It’s true that it works 95% of the time, maybe more. But as it sometimes fails, I personally don’t use it.
Right, good idea on cli update instead of gui. The first site is on Ubuntu 22.04, the second site that I still need to upgrade is on 24.04. I had enough excitement this AM, when I try the second site from cli update alone, I will report back. I still see the old vips on the second server /usr/local/lib/libvips.so.42, so hopefully this gets upgraded as needed this time, as it did not the first go around. Thanks
OJ just to be clear, my first discourse update definitely failed. even after a cli rebuild post gui update. Sooo, I had to manually update the containers to V18 of both Postgres and vips. I am hoping for a better outcome on the second site Stay tuned.
That is a pretty bad situation.
make sure you have enough disk space. Take note in the other topic that is recomended for 2 x database size
For sure, and I have plenty of disk space on both servers. This is server 1
df -h /var/discourse
Filesystem Size Used Avail Use% Mounted on
/dev/sda1 194G 25G 170G 13% /
Server 2:
df -h /var/discourse
Filesystem Size Used Avail Use% Mounted on
/dev/sda1 193G 26G 168G 14% /