[GitHub] [airflow] aquam8 edited a comment on issue #11979: Support helm install --wait for the chart

[GitBox <gi...@apache.org>]

aquam8 edited a comment on issue #11979:
URL: https://github.com/apache/airflow/issues/11979#issuecomment-831663761


   Hi,
   
   First of all, let me thank you for this project and your involvement in this awesome project.
   
   Now onto this issue.. I have come to realise that this is also the problem I am encountering.
   This is also the problem reported in this issue (see the use of --wait): https://github.com/apache/airflow/issues/15340 where it's reported that the webserver fails reporting no migrations ran.
   
   I could not understand why the Job resources would not get created, I could see the ServiceAccount resources but not the jobs.
   I then created YAML files manually from the helm get values command, and they got created and ran successfully.
   After looking more into it I have realised that they are Helm Hooks and indeed will not be installed when Helm is used in conjunction with --wait flag in an install/upgrade command.
   
   I would argue that you cannot make assumption on how helm update is run - usually it's as part of a CI/CD pipeline and applies to potentially many other helm releases. I don't think you can force people to forgot '--wait' flag especially because of its nature. As the OP explains, that flag is particularly critical as it waits for the upgrade to be successful or it rolls back.
   
   I think the solution is to not make these jobs as Hooks - but I'm not sure of the reason. In the past I have switched from Hooks to normal job to run my app migrations as part of normal app deployment. Applying migrations multiple time as no-op is a regular use-case.
   
   I would say that this is a serious issue and should be prioritised to allow a smooth install and upgrade. At the moment, fresh install for people using --wait will fail.
   
   
   Some more info on the problem:
   https://helm.sh/docs/topics/charts_hooks/ says:
   ```
   Note that if the --wait flag is set, the library will wait until all resources are in a ready state and will not run the post-install hook until they are ready.
   ```
   At the same time, the webserver & scheduler wait for the migrations to have run, so we have no READY pod, and it's a dead-end. Eventually the Helm --wait will time out and roll back.
   
   Thank you for your consideration
   
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org