Out of memory when running a lot of pods

[stijndehaes]

In environments where we are running a lot of pods we see the external resizer going out of memory.
In these environments we typically only run 10's of pvc's, but 1000's of pods.
I believe the pod informer might be the culprit.
Doesn't starting an informer launch a cache? This means all pods are currently cached on the resizer, which leads to excessive memory use. The code isn't actually using the cache, I think the code is only interested in the Add, Update, Delete part and not in the cache.

If this is correct I am willing to create a PR to fix this

[gnufied]

The whole handleInUse thing is behind a flag and only applicable for use cases where a driver needs external-resizer to enforce offline only expansion strongly. If it is enabled in EKS with external-resizer, then that seems like a problem. cc @wongma7

[stijndehaes]

@gnufied you are right, resizing can be done online on AWS i think. Good catch

[stijndehaes]

I will close this ticket as this was fixed in the EBS CSI driver, by setting the argument correctly