prometheus charts repo is invalid?

[crossplane-contrib]

prometheus charts repo is invalid?

[turkenh]

@jbw976 do you have following env var set in helm provider controller deployment? probably not?

    - name: XDG_CACHE_HOME
      value: /tmp

Probably this got lost during transition to new packaging.

[jbw976]

No I do not have that in my provider-helm deployment, I can try adding it and seeing if it fixes this @turkenh.

I would expect kubectl crossplane install provider crossplane/provider-helm:v0.3.0 to take care of this for me. Do we need to fix something (and where?) so that this works without manual intervention in v0.3.1 or something? 🙏

[turkenh]

I would expect kubectl crossplane install provider crossplane/provider-helm:v0.3.0 to take care of this for me. Do we need to fix something (and where?) so that this works without manual intervention in v0.3.1 or something? 🙏

Yes, this was already case with v0.2.0, but as stated, probably we lost this with new packaging changes. I remember, @hasheddan mentioned somewhere that this might got removed with new packaging (as we now use a static controller deployment).

Anyway, just pushed a possible fix (#37). It would be great if you could test it in your setup, built and pushed controller image as "turkenh/provider-helm-controller:fix36". Otherwise, I can test tomorrow.

[jbw976]

i tested with @turkenh fixed controller from above and it appears to have fixed the initial issue, but now I see this, is it supposed to be transient? its still there after 6-7 mins

  status:
    atProvider:
      releaseDescription: Initial install underway
      revision: 1
      state: pending-install
    conditions:
    - lastTransitionTime: "2020-10-20T23:53:15Z"
      message: 'update failed: failed to upgrade release: "platform-ref-aws-cluster-hmfxt-bt279"
        has no deployed releases'
      reason: ReconcileError
      status: "False"
      type: Synced

[turkenh]

'update failed: failed to upgrade release: "platform-ref-aws-cluster-hmfxt-bt279"
has no deployed releases'

This is because previous installation attempt got interrupted (helm controller pod got terminated for some reason) before it is completed and actual helm release is in pending-install state. There is a helm issue for this: helm/helm#7139

Just tested with the helm release you used in this issue and observed that chart has pre-install hooks for an initial job which means installation takes longer than usual, still should not be an issue if controller not stopped/killed.

So I wouldn't expect this would happen again with a clean installation.

[turkenh]

Still to overcome such an edge case:

We have an optional rollback support which rolls back latest release and retries which should also start supporting pending states with #37.

[turkenh]

@jbw976 crossplane/provider-helm:v0.3.3 has the fixes.

[jbw976]

Tested with v0.3.3 just now and it looks like this issue has been fixed!! thank you @turkenh!