ThePythonicCow experienced a failed Discourse upgrade on their forum, themooster.net, after running the standard ./launcher rebuild app command. The site returned a 502 Bad Gateway error, and logs revealed a PostgreSQL version mismatch: the data directory was initialized by PostgreSQL 13, but the system was running version 10.12. After consulting pfaffman’s link to the PostgreSQL 13 update guide, ThePythonicCow expressed frustration over the lack of clear warnings for casual users and confusion over whether the guide applied to their PostgreSQL 10.12 setup. After discovering their PG_VERSION file read “10”, they felt even more uncertain. Eventually, ThePythonicCow tried rebuilding again — despite ignoring most of the 2020-line output — and the site came back up successfully. pfaffman confirmed this is a known workaround: if the rebuild fails initially, retrying often resolves the issue. The takeaway: if ./launcher rebuild app doesn’t work the first time, try it again.
I have a little forum, themooster.net, based on Discourse that I had not upgraded in a couple of months.
I clicked “Update” in the “Admin” recommendation to upgrade, and it said I had to do the upgrade at the shell prompt instead, and gave me three lines to run:
cd /var/discourse
git pull
./launcher rebuild app
I ran those three commands. The “rebuild app” failed initially, because I didn’t have 5 GByte free on my little linode box, so I added more disk space and the “rebuild app” seemed to complete ok.
But the site didn’t come back up.
I looked around, rebooted, still not coming up, looked around a bit more.
Seems that the frontend nginx is coming up ok, but when I try to access my forum over the web, like a normal user, I get a " 502 Bad Gateway" error from the nginx frontend.
The file “/var/discourse/shared/standalone/log/var-log/postgres/current” is filled with errors such as:
2021-04-17 13:54:37.785 UTC [7095] FATAL: database files are incompatible with server
2021-04-17 13:54:37.785 UTC [7095] DETAIL: The data directory was initialized by PostgreSQL version 13, which is not compatible with this version 10.12 (Debian 10.12-2.pgdg100+1).
I have good backups of the forum in my directory:
/var/discourse/shared/standalone/backups/default
I’ll look at your PostgreSQL 13 link now - thanks.
Is there someway I should have, could have, known of this complication, without being a full time admin constantly following the activity of the key technologies (PostgreSQL, Ruby, Linux, …) behind Discourse?
If so, then could the 3 lines of instruction (cd /var/discouse; …) have another line “Warning: check here first for other information that might apply.”) ?
If not, then Discourse should be more open about it not being well suited for casual use.
So the above PostgreSQL 13 link, to the extent that it applies to a 12 ==> 13 upgrade, is NOT necessarily relevant, so far as I know (which is not far at all.)
I am having a really bad feeling about this … like trying to fix live military munitions, in a dark cave, without instructions, untrained, and not even being sure what kind of munitions I’m working on.
“You should rebuild again. My guess is that you ignored a message about the postgres upgrade.”
Got me thinking that I should try the “cd /var/discourse; git pull; ./launcher rebuild app” again … apparently that “rebuild” can be tried multiple times.
I captured the output this time. I GUARANTEE that I have ignored most of the messages in the output, however. There are 2020 lines of output from the “./launcher rebuild app” command alone. It would take me a month of Sundays to learn what all programs are running and what their output messages mean. There is no practical way for me to know if I am ignoring something critical in that output.
However …
It Worked !!
My website came back up, running the current version of software.
So the lesson for the day: If the “./launcher rebuild app” command fails to provide a working site, try it again.