High cpu usage and occasionall unavailability of metrics port about helm-exporter HOT 7 CLOSED

It will do it on each call to /metrics you can set it up to do this in the background and to cache the results if you use https://github.com/sstarcher/helm-exporter/blob/master/main.go#L50

from helm-exporter.

Thanks! I guess default prometheus scraping of 30s sometimes broke helm-exporter.
Much better now, might be worth better documenting that in helm chart? https://github.com/sstarcher/helm-exporter/blob/master/helm/values.yaml#L6

from helm-exporter.

It could certainly help people who have large clusters

from helm-exporter.

Hello!
I am having a similar problem.
My cluster has no Internet access.
I have set a variable "latestChartVersion: false", but have a long startup time and CrashLooping.

I tried setting a variable "initialDelaySeconds: 120" for reading and liveness probes, than I can see in logs
time="2021-12-10T16:06:27Z" level=warning msg="failed to update artifacthub cache due to Get "https://artifacthub.io/api/v1/helm-exporter\": dial tcp 13.33.240.37:443: i/o timeout "

Is there a standard way to disable artifacthub update?

from helm-exporter.

Your error is not the same thing. Your issue is it trying to fetch charts, but you have no internet access. I believe currently for it to work you would need to specify OverrideRegistries for each of your charts. If you do that it will not fall back to artifacthub.

Certainly a bug that it should not cause a crashback loop if it can't reach artifacthub

from helm-exporter.

Thanks for the quick response!

I have set the following settings, but I still get an unavailable artifacthub.
I don't have a real local helm repository with this charts.

helm-exporter:
...
  config:
    helmRegistries:
      override:
        - registry:
            url: "localhost"
          charts:
            - es
            - kibana
            - logstash
            - prometheus
            - ...

Why use artifacthub if I disabled latestChartVersion?

For info: When I set the setting "initialDelaySeconds: 120", the pod works fine, despite the warning.

from helm-exporter.

I bet the issue is it's failing in init where we do the cache.

So latest chart version is disabling the actual fetching, but we are still polling the cache. We likely need a way to disable the cache from happening or move it to a separate function call. I would consider that a bug.

from helm-exporter.