ocis icon indicating copy to clipboard operation
ocis copied to clipboard

ocis_full update collabora image results in startup failure

Open mmattel opened this issue 9 months ago • 4 comments

While testing to update images for traefik (3.3.3 --> 3.3.4, ok) and collabora, collabora fails to start. Tested on my Hetzner deployment.

Updating the collabora image from 24.04.12.3.1 to 24.04.12.4.1, a patch release which was published some days ago, the container does not start anymore and responds with the following cryptic error message. When switching back to 24.04.12.3.1, all works fine:

docker images

traefik                     v3.3.4         e49b83f3e049   22 hours ago   190MB
collabora/code              24.04.12.4.1   d3bfb5a571cb   5 days ago     1.43GB
collabora/code              24.04.12.3.1   e9fb61f587aa   2 weeks ago    1.43GB
onlyoffice/documentserver   8.3.0          497c96127bfc   2 weeks ago    3.54GB
traefik                     v3.3.3         1c768f87626a   3 weeks ago    190MB
owncloud/ocis               latest         ccb91a2b7514   2 months ago   186MB
owncloud/ocis-rolling       7.0.0          e0688db571c1   2 months ago   186MB
owncloud/ocis-rolling       latest         00564f82b13f   4 months ago   185MB
apache/tika                 latest-full    6c5f30726faa   4 months ago   733MB

docker compose ps -a --format "table {{.Service}}\t{{.State}}\t{{.ID}}"

SERVICE            STATE        CONTAINER ID
collabora          restarting   eefedafbef2d
collaboration      created      e36b13edb49d
collaboration-oo   running      4aab9af30c59
ocis               running      a68251298c39
onlyoffice         running      1d5131f9d6c1
tika               running      bb3bd87c8190
traefik            running      8e8fcd95cb24

docker logs eefedafbef2d 2>&1 | less

Unknown option specified: c

Note that we have for a good reason a defined image for traefik, collabora and onlyoffice to avoid such scenarios in production, the currently used images as shown above deploy well.

@kobergj @2403905 we need to investigate after 7.1

mmattel avatar Feb 26 '25 08:02 mmattel

The error comes from the command: ["bash", "-c", "coolconfig generate-proof-key ; /start-collabora-online.sh"] https://github.com/owncloud/ocis/blob/c954c9064c57ccd30a8baf1c3da8a36977374484/deployments/examples/ocis_full/collabora.yml#L83

2403905 avatar Feb 26 '25 17:02 2403905

Just tested with the new Collabora Code version: 24.04.13.1.1 --> sadly the same result. The container errors when starting.

mmattel avatar Mar 05 '25 17:03 mmattel

Creating containers with working/not working images, the following shows up. Running the command coolconfig generate-proof-key in the containers shell produces different results:

  1. With working images, the command succeeds. The container is still up and running after the command.
  2. With non working images, the command fails. The container closes like when you shut it down. The bad thing is, that you cant see if there is an error because the container just crashed, there is no output

mmattel avatar Mar 06 '25 11:03 mmattel

Perhaps the change was https://github.com/CollaboraOnline/online/pull/11177 which has:

-CMD ["/start-collabora-online.sh"]
+ENTRYPOINT ["/start-collabora-online.sh"]

$ docker run -it collabora/code:24.04.12.3.1 bash -c "coolconfig generate-proof-key ; echo hello world" hello world

vs

$ docker run -it collabora/code:24.04.12.4.1 bash -c "coolconfig generate-proof-key ; echo hello world" Certificate request self-signature ok subject=C = DE, ST = BW, L = Stuttgart, O = Dummy Authority, CN = localhost Unknown option specified: c

while disabling the entrypoint like below succeeds again

$ docker run --entrypoint "" -it collabora/code:24.04.12.4.1 bash -c "coolconfig generate-proof-key ; echo hello world" hello world

so maybe something along the line of

entrypoint: [""]

in your collabora.yml gets you to where you want to be

caolanm avatar Mar 31 '25 12:03 caolanm