Helm Provider: rpc error (timeout) during update of release about keel

[rusenask]

created new release with 0.4.1, cloud.docker.com is building it now (although it's quite slow) so image should be available soon. Configurable timeout could be in values.yml with other config:

keel:
  # keel policy (all/major/minor/patch/force)
  policy: all
  # update timeout in seconds (defaults to 300)
  timeout: 9001
  # images to track and update
  images:
    - repository: image.repository # [1]
      tag: image.tag  # [2]

Would that make sense?

[fujin]

confirming fixed in 0.4.1:

[rusenask]

okay, 0.4.1 is available, just noticed that my own Keel upgraded itself and sent notification to slack faster than I got an email from dockerhub. Hope that helps!

[rusenask]

Hi! there is 30 seconds timeout for release upgrades, I guess it's a bit optimistic if you have larger images?

[fujin]

@rusenask yea, that could well be it. this image is 625MB (has a lot of runtime tooling) and there are a number of replicas in the deployment (a HPA controls it, so it varies, but anywhere from 2~36)

gcr.io/dd-engineering/identity 1.8.1 69916c4eff80 7 hours ago 625MB

the ability to control the timeout via configuration would be welcome as a workaround for now!

[fujin]

Is that timeout @ https://github.com/rusenask/keel/blob/28553b3e9db8076de8c5b0d50eb664c4b6f7e92f/provider/helm/helm.go#L329 ?

[rusenask]

Yeah that's the one. Are you modifying timeout when updating releases with CLI? I see default 300 seconds in helm. I can quickly bump it and add configuration later.

[fujin]

@rusenask we have not had to modify the timeout when updating releases with CLI, no.

that sounds awesome to me. 🥇 the Helm stuff looks great so far.

[theckman]

@rusenask (I work with @fujin) -- that would definitely work for us!

[fujin]

@rusenask yup, looks good to me. 👍

[rusenask]

I guess it would be nice to see how long release update took, from time it looks like ~1 minute :)

P.S. Bonus feature - if you invite Keel to your channel (or you can message it directly) you can:

[theckman]

@rusenask for the latest test deployment, webhooks were 47 seconds apart.

[theckman]

I guess I should mention that this is our smallest environment with only two Pods in the deployment. Other environments may have 6+.

[rusenask]

That timeout is for single helm deployment, each deployment will have a refreshed timeout. Since Helm provider is slightly different from k8s - it would make sense to start helm upgrades in parallel (as it's waiting for ok/not ok) .