Collabora Online stops working after restart (502 Bad Gateway)

[joseph]

@archos you can try visiting https://collab.example.com/hosting/discovery on the browser . It should display some xml in the browser. If it works, then try curl of that URL from web terminal of nextcloud. Does that work? (Maybe there is cloudflare or some proxy in the middle blocking requests?)

[archos]

@joseph I tried opening https://collab.example.com/hosting/discovery in a browser, and I got the same page as before ("This app is currently not responding").

When I try it from the Nextcloud terminal using curl, I receive the HTML code of the Cloudron App Error page, which includes styling and the message:

"This app is currently not responding. Please try refreshing the page in a few minutes."

What should I check next?

Thank you for your response. [link text](link url)

[joseph]

OK.. Not sure. Does this happen with a fresh install? I would say check if reinstall works because the app itself is stateless (has no data) . Collabora updates are quite hard to debug since they don't provider a real changelog

[archos]

@joseph Thank you for your response.
I have tried reinstalling the application multiple times. I even installed it on a different domain once.

But after every restart, the issue remains the same.

[joseph]

@archos is there any proxy/cloudflare or anything in the network we need to be aware of?

[joseph]

Just installed in https://collabora.demo.cloudron.io/hosting/discovery and fresh install works atleast

[nebulon]

@archos do you see anything in the collabora app instance logs? Does it even startup fine? Since you get the default cloudron error page as a result, it seems that the app itself (in the container) isn't healthy for a start.

[archos]

@nebulon The application seems to work normally right after installation.
I tried connecting two Nextcloud instances – one on the same server as Collabora, and another on a different server.

When I add the "Memory Limit" application, after a restart, Collabora stops working, and I can no longer connect.

I also tried keeping only the instance that is on the same server as Collabora, but it only works until the next restart.

I’m not sure if it’s related, but the only errors I found in the logs are:

2025-03-17T20:08:05.532Z box:dns/cloudflare upsert: Adding new record fqdn: _acme-challenge.collab.example.com, zoneName: example.com proxied: false
2025-03-17T20:08:07.644Z box:dns/waitfordns isChangeSynced: NS damon.ns.cloudflare.com errored when resolving _acme-challenge.collab.example.com (TXT): Error: queryTxt ENOTFOUND _acme-challenge.collab.example.com
2025-03-17T21:11:21+01:00 Failed to initialize COOLWSD: SSL context exception: Error loading private key from file /etc/coolwsd/key.pem: error:8000000D:system library::Permission denied

[nebulon]

What is a "Memory Limit" application in this case?

Not sure what that .pem file is for. There seems to exist one from the initial package:

root@3c01b922-d875-4bc1-99b9-20c99a0d50c5:/app/code# ll /etc/coolwsd/key.pem
-rw------- 1 root root 1704 Mar 14 05:22 /etc/coolwsd/key.pem

but I can't think of how this would be used in Cloudron, since the reverse proxy with SSL termination is not inside the container in Cloudron.

[archos]

@nebulonThe "Memory Limit" setting in this case refers to the RAM allocation for Collabora Online in the Cloudron Dashboard.
Initially, the application was running with 1GB of RAM, but I increased it to 3GB to see if it helps with stability.

The only relevant issue I found is this one
but I’m not sure if it’s the same problem.

Unfortunately, I don’t have much experience with this.

[nebulon]

Right the linked nextcloud forum entry is about the reverse proxy which in Cloudron's case is part of the platform. So you see no errors in the app logs itself and it starts up fine from the logs, but the healthcheck just doesn't succeed?

[archos]

@nebulon After restarting Collabora Online, I checked the logs and found the following errors:

connect() failed (111: Connection refused) while connecting to upstream, client: 172.18.0.1, server: , request: "GET /hosting/discovery HTTP/1.1", upstream: "http://127.0.0.1:9980/hosting/discovery"
Failed to initialize COOLWSD: SSL context exception: Error loading private key from file /etc/coolwsd/key.pem: error:8000000D:system library::Permission denied
sh: 1: /usr/bin/coolmount: Operation not permitted

It seems like Collabora Online fails to properly start after a reboot. Could the SSL certificate issue or permission errors be causing this? Any suggestions on what to check next?

[nebulon]

very strange that I cannot reproduce this anywhere. Can't see why it would try to load some private key in the first place. Did you tweak any of the configs?

[archos]

@nebulon No, I haven’t manually modified any configuration files. Everything was set up through the Cloudron interface.

I have tried reinstalling Collabora multiple times, even on a different domain, but the issue persists. After installation, it works fine, but after a restart, it stops responding.

The only changes I made were increasing the memory allocation and connecting multiple Nextcloud instances.

[nebulon]

I have tried to reproduce this with two nextcloud instances having office documents open and restarting collabora various times. Unfortunately I cannot reproduce any of those bits. Are you able to reproduce this on the demo cloudron?

[archos]

@nebulon Hi,
I tried the demo instance of Collabora, but after a restart, I get the same error. I must be doing something wrong, but I really don't know what.

I logged into Collabora and added my two Nextcloud domains: app1.demo.cloudron.io|app2.demo.cloudron.io. After a restart, I can no longer connect.

At least now we know that the issue is somewhere on my side.

[archos]

I also tried a fresh Nextcloud installation. My first Nextcloud instance is on the main domain, so I thought that might be the issue, but even with the new instance I have the same problem.

After restarting the Collabora app, the same error appears again.

It seems like someone had a similar issue here

[nebulon]

Aha, so possibly then changing the domain from default could be causing the issue. I didn't attempt that, since the test nextcloud instances are on the same main domain. Lets see if I can then reproduce it...

[nebulon]

I can now reproduce this by setting explicit domains. We have to debug and hopefully fix the package then. For the time being whildcard subdomains curiously seem to work fine as a workaround.

[msbt]

@nebulon can you also test it it multiple domains? I remember having issues with that a while back