-
What's your question or issue?I recently noticed that my file tasks queue wasn't being processed and upon investigation, I found that under "System Status" "Celery Status" was reporting "Error". This is in the What have you tried?What have I tried:
About my setup:
One thing that did seem to make a little impact was manually running celery from the container CLI: This would start processing documents, it would move duplicates into the "failed" status, but as soon as it hits one that isn't already present, it would sit in the "started" status for 30 minutes and then time out. But then "Celery Status" under "System Status" would change to OK Paperless-ngx version2.2.15 Host OSKubernetes Installation methodDocker - official image System status{
"pngx_version": "2.20.15",
"server_os": "Linux-6.18.32-0-lts-x86_64-with-glibc2.41",
"install_type": "kubernetes",
"storage": {
"total": 580139352064,
"available": 379011989504
},
"database": {
"type": "mysql",
"url": "inftechPaperless",
"status": "OK",
"error": null,
"migration_status": {
"latest_migration": "documents.1075_workflowaction_order",
"unapplied_migrations": []
}
},
"tasks": {
"redis_url": "redis://localhost:6379",
"redis_status": "OK",
"redis_error": null,
"celery_status": "ERROR",
"celery_url": null,
"celery_error": "Error connecting to celery, check logs for more detail.",
"index_status": "OK",
"index_last_modified": "2026-06-04T10:09:33.034878Z",
"index_error": null,
"classifier_status": "OK",
"classifier_last_trained": "2026-05-25T18:05:03.489469Z",
"classifier_error": null,
"sanity_check_status": "OK",
"sanity_check_last_run": "2026-06-04T08:40:16.851048Z",
"sanity_check_error": null
},
"websocket_connected": "OK"
}Relevant logs or outputNo response |
Beta Was this translation helpful? Give feedback.
Replies: 7 comments 7 replies
-
|
Unfortunately, that error under system status is misleading in that there may not really be a problem with celery itself—it's on my list of things to hunt down. But as for the actual tasks not running, are there errors about that or you just noticed that they havent run recently? Can you check the redis logs? Also are you sharing the redis instance between two installs (that has been problematic in the past)? |
Beta Was this translation helpful? Give feedback.
-
|
I'd be happy to help with hunting down. I have already gone down quite a rabbit hole with them, but my lack of understanding how it is "supposed" to be working is holding me back from making any definitive conclusions. One thing that might help with that would be making The tasks not running under "Health" is something I didn't notice until I started troubleshooting this. My main issue is with the "File Tasks" not being processed. Both of my instances seem to have broken around May 25. The runs of the search index and sanity checker on 04 Jun were manually started. On my other instance (where I have not done any more troubleshooting besides concluding that it is also broken), all three have last run times on May 25. The Redis logs do not seem to give any clues, I just restarted my setup to get it back to the 'normal' state, where I haven't fiddled with anything in the container itself. Both instances have their own Redis instance and their own Mariadb Database and their own data directory and run in separate namespaces with different domain names, they are not aware of each other. |
Beta Was this translation helpful? Give feedback.
-
|
Yea I see you've tried the usual suspects, removing Certainly all having the same date is suspicious for something about the deployment changing around then. The fact that manually starting a Celery worker causes documents to start processing and makes System Status report OK makes me think the worker service may not be running correctly, or it is starting and then failing. Could you check inside the container whether the celery worker and scheduler processes are present?
and Also the container startup logs might be helpful, not just paperless.log Lastly, the data over NFS isnt something that changed? |
Beta Was this translation helpful? Give feedback.
-
And container logs: And, no, data over NFS has been the case ever since I started using Paperless (and is actually a pattern I use with all my containers) |
Beta Was this translation helpful? Give feedback.
-
|
Weird, so it's up. Actually maybe inspecting the queue: and you can do the same with (run from inside the container) |
Beta Was this translation helpful? Give feedback.
-
|
All four have the same response: |
Beta Was this translation helpful? Give feedback.
-
|
At the risk of being not helpful, I ran this by Codex:
|
Beta Was this translation helpful? Give feedback.
-
(hangs "forever", no output) My last Redis update was on May 27, so just after the last success, so I downgraded to 8.6.3 and then to 8.6.2, deleted |
Beta Was this translation helpful? Give feedback.
-
|
Hmm so the Celery worker is hanging during startup, before it becomes an actual Celery node. Again ChatGPT:
|
Beta Was this translation helpful? Give feedback.
-
|
OK. Progress! That just leaves me with one puzzle... Why did this work for years and stopped now... Thanks so much for helping out with this! |
Beta Was this translation helpful? Give feedback.
-
|
Glad it’s solved but I know better than to ask questions like “why” when computers decide to act up 😅 Yea I dunno, maybe I’ll run it by an LLM to see if there’s anything to do on our side but it sounds like a k8s or NFS kinda mystery |
Beta Was this translation helpful? Give feedback.
-
|
It seems a bit specific (not aware of a lot suppose if theres a lesson here its:
Again, perhaps if this becomes a more frequent issue... |
Beta Was this translation helpful? Give feedback.
-
|
So, today I wanted to fix my other instance, but because I now knew that the issue had something to do with NFS, I decided to check if something had changed on my NFS server around the moment it broke, and lo and behold, I had run updates on it right around the time it broke. Checking the logs, I found, several messages in the logs: Not sure why only this server had this problem, but I gave the nfs daemon a restart... and what do you know... that fixed it on the other instance (logs litterally changing from 27-05 to 17-06... the process had hung for 3 weeks and now it just continued like nothing happened). |
Beta Was this translation helpful? Give feedback.
-
|
This discussion has been automatically closed because it was marked as answered. Please see our contributing guidelines for more details. |
Beta Was this translation helpful? Give feedback.
Hmm so the Celery worker is hanging during startup, before it becomes an actual Celery node. Again ChatGPT: