Config reload without restarting pods

[holmesb]

This is a followup to #258 which proposes restarting pods (using kubectl rollout restart) to reload config. The problem with this is there is downtime while pods restart.

It would be better to use a sidecar which watches for configmap changes and then sends fluentd's /api/config.gracefulReload command. If no PR materialises, I'll create one when I get time. I've done similar for other charts (eg Jenkins\Grafana).

(this issue will become redundant if this feature lands in fluentd - but doesn't look imminent)

[mwang2016]

Hi @holmesb thank you for submitting this enhancement request. This is a good idea; if you can submit a PR for this, it will be very helpful.

[vinzent]

@mwang2016 I would like to create a PR for that config reload feature.

there is the https://hub.docker.com/r/jimmidyson/configmap-reload image which OpenShift uses (or uses to builds its own https:/openshift/configmap-reload)

is this image acceptable?

[vinzent]

@mwang2016 ?

[holmesb]

The more widely used the sidecar image is, the more secure/stable/better. Also helps if the dockerfile used to build is available, and there's a clear link between the docker image version and the code version used to create it.

The main Jenkins helm chart in Cloudbees' repo uses kiwigrid/k8s-sidecar, which has 10m downloads, and has worked well for me in the past. It ticks these three boxes. Openshift/configmap-reload ticks the first two, but not sure how the docker image version (v5.0) relates to a release on github (v0.2.2). They seem to have similar functionality though.

[vinzent]

https://hub.docker.com/r/jimmidyson/configmap-reload is not built from https:/openshift/configmap-reload . OpenShift forked the repo to build its own. OpenShift has its own private registry (registry.redhat.io). The image source for jimmidyson/configmap-reload is here: https:/jimmidyson/configmap-reload .

On Docker Hub the jimmidyson/configmap-reload has 50m downloads.

[github-actions]

This issue is stale because it has been open for 30 days with no activity.

[github-actions]

This issue is stale because it has been open for 30 days with no activity.

[github-actions]

This issue was closed because it has been inactive for 14 days since being marked as stale.