User Godmar_Back reports a 502 Bad Gateway error after upgrading Discourse via the admin interface. Inside the container, nginx expects a server at localhost:3000, but the Unicorn process fails to start, repeatedly logging “No such process” errors. They seek debugging guidance on what command Discourse is attempting to run. merefield suggests checking a similar past issue and asks if the install is vanilla; Godmar_Back confirms using plugins (including data-explorer) and running behind host nginx. david suggests the issue may stem from a recent reverted data-explorer change and advises rebuilding. Godmar_Back tries git pull and ./launcher restart app, then ./launcher rebuild app, but the issue persists. merefield confirms the database is safe during rebuild and notes that restart after git pull is insufficient. Godmar_Back asks if rebuild updates plugins; merefield confirms it does, if plugins are cloned in app.yml. DonH shares a similar experience: a 502 occurred after system updates (apt upgrade), not directly after Discourse upgrade, and was resolved by ./launcher rebuild app, which also fully updated Discourse. Plugins in use include discourse-data-explorer, discourse-bbcode, discourse-docs, docker_manager, and styleguide.
After recently using the admin upgrade interface, my discourse instance stopped working, responding with 502 Bad Gateway.
I’ve entered the container and it appears to be running an nginx that is expecting a server at localhost:3000, which is not running.
(54) Waiting for new unicorn workers under 3802725 to start up...
(54) Waiting for new unicorn workers under 3802725 to start up...
(54) Old pid is: 3800363 New pid is: 3802725
config/unicorn_launcher: line 71: kill: (3802725) - No such process
config/unicorn_launcher: line 15: kill: (3802725) - No such process
(54) exiting
ok: run: redis: (pid 62) 3418739s
ok: run: postgres: (pid 53) 3418739s
supervisor pid: 3803896 unicorn pid: 3803900
config/unicorn_launcher: line 71: kill: (3803900) - No such process
config/unicorn_launcher: line 15: kill: (3803900) - No such process
(3803896) exiting
this is followed repeatedly by:
ok: run: redis: (pid 64) 4905s
ok: run: postgres: (pid 65) 4905s
supervisor pid: 18571 unicorn pid: 18575
config/unicorn_launcher: line 71: kill: (18575) - No such process
config/unicorn_launcher: line 15: kill: (18575) - No such process
(18571) exiting
I’d like to start this thread for help in debugging this; what’s the next step here - which command is Discourse trying to run. (I know I could find this out by reading/reverse engineering the code, but it may be useful to have a thread on this on the forum.)
Given the timing, this is most likely related to a data-explorer change which caused some issues. We’ve now reverted it, so if you try the rebuild again it should work better
Yes, I’m using data explorer. I didn’t do a git pull before restarting.
When I do a git pull, and then ./launcher restart app it’s not fixed.
Except that I’m running it behind an nginx on the host.
(And I have a few plugins, such as data explorer.)
I’m now trying ./launcher rebuild app - I hope that rebuilding the app will preserve my forum’s database … and I’m not ending up with my forum reset.
Doing launcher rebuild app does not address the issue.
This post describes an issue with privileged vs unprivileged containers, but doesn’t provide more information. It’s also from 2 years ago so may not be related to a recent update.
In case this is still being investigated I had a 502 gateway error but not directly after the update routine which failed midway with a Ruby versioning error. Since I had not updated the server in about six weeks I ran apt update/upgrade and rebooted. That’s when the 502 error occurred, I could not raise the forum website. Rebuilding app fixed things and also updated Discourse fully.
For the record I have these plugins installed and enabled:
