search

fabric8io / fluent-plugin-kubernetes_metadata_filter

Enrich your fluentd events with Kubernetes metadata
Apache License 2.0
350 stars 166 forks source link

Reduce kubeclient timeouts to reasonable values #350

Closed orgads closed 2 years ago

orgads commented 2 years ago

The API should be very responsive. If it hangs, it is usually because something is wrong, and waiting longer won't help.

orgads commented 2 years ago

We had a pod that got a reply for get_pod, but timed out (60s) on get_namespace. Ideally, I'd prefer to retry 2-3 times on such cases. What do you say?

What's the preferred way to retry?

jcantrill commented 2 years ago

We had a pod that got a reply for get_pod, but timed out (60s) on get_namespace. Ideally, I'd prefer to retry 2-3 times on such cases. What do you say?

This may be true but what about cases with networks which have high latency? You may begin to have throughput issues but retrying won't help in these cases.

What's the preferred way to retry?

I believe the better way to resolve this issue is to introduce a configuration parameter that allows user's to tweak this setting instead of assuming the defaults provided here fit everyones use case. These may be fine for defaults but we should also be able to override them.

orgads commented 2 years ago

Done (no retry yet).

jcantrill commented 2 years ago

Looks good. Please also update the README

orgads commented 2 years ago

Looks good. Please also update the README

Thank you, done.

jcantrill commented 2 years ago

https://rubygems.org/gems/fluent-plugin-kubernetes_metadata_filter/versions/3.1.0

orgads commented 2 years ago

Thank you