Helm Releases fail with - HelmChart ' ' is not ready

[vdonadze]

When I apply helm releases to the cluster (10-15 of them), initial reconciliation for each of them fails with HelmChart ' ' is not ready message. After some time second reconciliation starts and all of them become successful and ready.
Could you please advice how can I avoid failing on initial reconciliation ?

I tried disableWait for each section, but no result.

spec:
  chart:
    ....
  install:
    createNamespace: false
    disableHooks: true
    disableOpenAPIValidation: true
    disableWait: true
    disableWaitForJobs: true
  interval: 200m
  maxHistory: 2
  timeout: 2m
  uninstall:
    disableHooks: true
    disableWait: true
    keepHistory: false
  upgrade:
    disableHooks: true
    disableOpenAPIValidation: true
    disableWait: true
    disableWaitForJobs: true
    force: true
    preserveValues: false

kubectl get hr (initial reconciliation, when helm releases were just created)
lw-data-provider-http   41s   False     HelmChart 'service/th2-demo-books-lw-data-provider-http' is not ready
mstore                  41s   False     HelmChart 'service/th2-demo-books-mstore' is not ready
rpt-data-provider       42s   False     HelmChart 'service/th2-demo-books-rpt-data-provider' is not ready
rpt-data-viewer         41s   False     HelmChart 'service/th2-demo-books-rpt-data-viewer' is not ready
script                  40s   False     HelmChart 'service/th2-demo-books-script' is not ready

kubectl get hr (second reconciliation, happens automatically after 1-2 minutes)
message-producer        12m   True    Release reconciliation succeeded
mstore                  12m   True    Release reconciliation succeeded
rpt-data-provider       12m   True    Release reconciliation succeeded
rpt-data-viewer         12m   True    Release reconciliation succeeded
script                  12m   True    Release reconciliation succeeded

[kingdonb]

I think this is normal and expected, although it is annoying, the first time the HelmRelease is reconciled it is indeed not ready. It has to write a HelmChart to the Kube API, which is then reconciled by source controller, and it can't do anything until after that happens.

I'm not sure if there's a way to make this behavior better, because I understand what you're experiencing and I have it too. HelmRelease needs to be reconciled twice before the install will succeed, it can be done by interval or by CLI, or with any of the interfaces for Flux like Weave GitOps, the VSCode GitOps Tools Extension, or any other tool that can run flux reconcile hr.

This is one of the pains that is definitely made slightly less intense by the presence of a red X in your UI that you can click and kick. I'm not usually a big UI proponent, preferring to operate the CLI most of the time, but this is one place it really can come in handy.

[vdonadze]

Thanks for answering. Now that I know that this is expected behavior I can start working on my custom solution. I was worried that I had something configured wrong.

[kingdonb]

One of these settings may help with the requeue timing:

https://fluxcd.io/flux/cheatsheets/bootstrap/#increase-the-number-of-workers

I think it might just be a problem with Helm controller and long intervals. We should probably consider it a bug if it doesn't requeue with some delay after sending a HelmChart to the source controller for reconciling, or maybe some other mitigation can be applied?

[Eforen]

is there a way to get from flux why the Chart is not ready? I tried the flux reconcile hr command and it didn't fix the problem I am having.

[stefanprodan]

See https://fluxcd.io/flux/cheatsheets/troubleshooting/#how-to-debug-not-ready-errors