Suggestion: Use Horizontal Pod Autoscaler to manage cluster size about fluent HOT 2 CLOSED

Thanks for all the suggestions again, @Waterdrips. I'm not sure this one will work for our purposes for a couple reasons.

1. According to the link you shared above, the horizontal pod autoscaler "... does not apply to objects that can’t be scaled, for example, DaemonSets.".
2. Our policy for autoscaling relies on a combination of a variety of metrics that are mostly collected internally (e.g., key access frequencies, client-reported e2e request latencies, etc.), which the HPA doesn't seem to have an API for dealing with.

That said, now that I'm moving to deploying application components with DaemonSets, the scaling operations become a lot easier -- we just need to change the size of the kops instance groups, and the DaemonSet takes care of the rest.

Closing this issue for now -- feel free to reopen if there's something I'm missing here!

from fluent.

@vsreekanti - I was thinking of the cluster autoscaler - it wont work if deployments are moving to DaemonStets also.

from fluent.