Cannot connect to immich_machine_learning #11062
Replies: 6 comments 4 replies
-
|
That env var should still work, but I don't see it in your .env. You can also configure the URL from the admin panel: https://my.immich.app/admin/system-settings?isOpen=machine-learning |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Github has some problem with email notifications. I haven't noticed your comment. @bo0tzz env var didn't work I tried them all. env file aren't used in quadlet files. But the variables can be put there as environment variables. Connection to postgress, redis etc I have maintained just like that. When I changed it via admin panel to "http://10.0.0.5:3003", it shown different message. And this is from immich machine learning: I don't have any environment variables in immich_machine_learning container. There is .env file in docker compose, but it doesn't seem any variable in .env is needed for this container. |
Beta Was this translation helpful? Give feedback.
2 replies
-
|
That's a DNS issue, your ml container is not able to reach dns (or the internet at all, potentially) to download the models. |
Beta Was this translation helpful? Give feedback.
-
|
Server has access to internet. I'm not blocking access to internet from container. SeLinux is in permissive mode, DNS translation from server works. I don't know exactly how can I verify it from the immich container. It doesn't seem to have ping installed or have an access to global repositories. |
Beta Was this translation helpful? Give feedback.
-
|
Can it be caused by anything else? Which address is immich even checking? |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Beta Was this translation helpful? Give feedback.
-
|
Hmm, But nothing seems to be solution I could do my own. Do you have some running immich_machine_learning, which is working right now? |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Yes, the machine learning code in general works fine. This really is an issue with your network somewhere, which I don't have any insight into. |
Beta Was this translation helpful? Give feedback.
-
|
hmm, it seems issue is in podman network. I will contact them. That's all. Thanks for your help. |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
versions:
I'm experiencing connection troubles between the server and the machine learning container as well. which is a bit weird as:
I can reproduce the errors by: Any suggestion? I've 22k+ pictures in my library, idk if that's lots or a few for immich. I plan to add some GPUS nodes to my cluster, but having all intel processors I'm hoping |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
The bug
I cannot connect to immich_machine_learning
I'm using quadlet podman and I have all containers separated in virtual lan with ip adresses.
In past, I think I could handle it with parameter IMMICH_MACHINE_LEARNING_URL, eg:
What is currently replacement for this environment variable? I'm not sure if immich-machine-learning is some alias for localhost, but it will not work on lan.
The OS that Immich Server is running on
podman in fedora server
Version of Immich Server
1.0.8
Version of Immich Mobile App
1.0.8
Platform with the issue
Your docker-compose.yml content
Not used.Your .env content
Reproduction steps
1.It seems machine learning is used only during upload 2. so i upload something and seen there are errors in log. 3. ...Relevant log output
No response
Additional information
No response
Beta Was this translation helpful? Give feedback.
All reactions