search

langflow-ai / langflow-helm-charts

Langflow Helm Charts
MIT License
21 stars 12 forks source link

Helm Chart deployment #10

Closed whur closed 2 months ago

whur commented 3 months ago

Can you update the helm chart deployment and version bump? I'm still getting after #8 with a fresh install i just started today:

Helm install failed for release default/langflow-ide with chart langflow-ide@0.0.1: template: langflow-ide/templates/backend-statefulset.yaml:148:35: executing "langflow-ide/templates/backend-statefulset.yaml" at <.Values.zookeeper.volumes.data.existingStorageClassName>: nil pointer evaluating interface {}.volumes

Please and thank you.

hakan-77 commented 3 months ago

Versions after 1.0-alpha have some significant changes. Any chance to upgrade soon? or, are we going to wait 1.0.beta?

BTW, in my case (1.0.11) everything installs fine but the frontend cannot reach backend. Error message in the UI only, no backend logs.

nicoloboschi commented 3 months ago

Hey @whur and @hakan-77 I'm releasing 0.1.0 for both the charts.

There's still a bug for the frontend in the ´ide´ chart that needs a fix on the docker image side. https://github.com/langflow-ai/langflow/pull/2887

whur commented 3 months ago

Does this health check PR fix this issue?

Capture

I can't find anything in either pod's logs that identify what's going on.

hakan-77 commented 3 months ago

@nicoloboschi sweet! @whur I face the exact same issue, re: no log. I think we have to change log level from critical to error. I see this is done for runtime via env, I think we can do the same for ide (will try this soon) https://github.com/search?q=repo%3Alangflow-ai%2Flangflow-helm-charts%20LANGFLOW_LOG_LEVEL&type=code

nicoloboschi commented 3 months ago

The specific error is that frontend is checking whether the backend is up or not by calling the health check. Since that endpoint is not reachable it just gets stuck. The request doesn't arrive to the backend at all, so there's no log

ragingbal commented 3 months ago

Same issue here with the latest charts on GKE. The backend pod in a crashloop without much in terms of logs.

Any ETA for fix?

whur commented 3 months ago

I'm still getting health check errors on the langflow backend:

  Normal   Pulled     17s   kubelet            Container image "langflowai/langflow-backend:1.0.13" already present on machine
  Normal   Created    17s   kubelet            Created container langflow-ide
  Normal   Started    17s   kubelet            Started container langflow-ide
  Warning  Unhealthy  3s    kubelet            Liveness probe failed: Get "http://10.69.1.174:7860/health": context deadline exceeded (Client.Timeout exceeded while awaiting headers)
  Warning  Unhealthy  3s    kubelet            Readiness probe failed: Get "http://10.69.1.174:7860/health": context deadline exceeded (Client.Timeout exceeded while awaiting headers)

i also have this from the logs, this is a fresh install using postgres:

Starting Langflow v1.0.13...
[07/28/24 06:58:36] ERROR    2024-07-28 06:58:36 - ERROR    - utils  utils.py:27
                             - Error retrieving messages from                   
                             monitor service: Catalog Error: Table              
                             with name messages does not exist!                 
                             Did you mean                                       
                             "temp.information_schema.tables"?                  
                             LINE 1: ...ession_id, text, files,                 
                             timestamp FROM messages ORDER BY                   
                             timestamp DESC                                     

                                        ^                                       
                    ERROR    2024-07-28 06:58:36 - ERROR    -     service.py:221
                             service - Error migrating                          
                             transactions from monitor service to               
                             database: Catalog Error: Table with                
                             name transactions does not exist!                  
                             Did you mean                                       
                             "temp.information_schema.columns"?                 
                             LINE 1: ...ex_id, inputs, outputs,                 
                             target_id FROM transactions LIMIT                  
                             100
nicoloboschi commented 3 months ago

Which version are you using?

whur commented 3 months ago

I'm using langflow 1.0.13 and helm chart 0.1.0

whur commented 3 months ago

Langflow seems to be working and the health checks may be an environment issue on my side, so I think we can close this.

tkdguq05 commented 3 months ago

@whur Did upgrading to version 1.0.13 solve the couldn't estabilish a connect issue? I had a same problem on 1.0.11.

tkdguq05 commented 3 months ago

version 1.0.14 works fine!

whur commented 3 months ago

@tkdguq05 no, I still get the health check issues but i have some health check errors on other pods so I can't rule out environment issue at the moment, and I haven't had the time to get into that.

hakan-77 commented 2 months ago

version 1.0.14 works fine for me too, except, frontend to backend communication with http did not work. I had to use https end to end. FYI- @whur

nicoloboschi commented 2 months ago

@whur @hakan-77 did you have a chance to test this out with 1.0.16? I'd like to close this issue

whur commented 2 months ago

@whur @hakan-77 did you have a chance to test this out with 1.0.16? I'd like to close this issue

i have not, but my issues may be environmental so this can be closed.

hakan-77 commented 2 months ago

@nicoloboschi I have tested with both 1.0.16 and 1.0.17. Everything works, thank you for the great chart!