Performance problem about pod informer

[hykych]

	// Create pod informer.
	podInformer := kubeInformerFactory.Core().V1().Pods()

	// Set up an event handler for when pod resources change
	podInformer.Informer().AddEventHandler(cache.ResourceEventHandlerFuncs{
		AddFunc:    jc.AddPod,
		UpdateFunc: jc.UpdatePod,
		DeleteFunc: jc.DeletePod,
	})

I've seen this informer in the code. Doest it have performance problem since it doesn't filter any kind of pod, when the replicas of tf-operator become larger the api-server will send the same event to each one of tf-operator which brings a lot of pressure to api-server and etcd.

[issue-label-bot]

Issue Label Bot is not confident enough to auto-label this issue. See dashboard for more details.

[gaocegege]

Thanks for your issue. If you mean FilteringResourceEventHandler, it applies the provided filter to all events coming in. Thus it does not help us solve the performance issue. many controllers in Kubernetes, such as job_controller, also use ResourceEventHandlerFuncs.

And we have filter logic in jc.AddPod, thus the operator will avoid redundant reconcile.

Welcome any suggestion and help on it.

[hykych]

@gaocegege so do you guys have any plan or ideas to solve this? or have you benchmarked to show how severe the problem is?

[gaocegege]

Benchmark is here #829

[gaocegege]

@hykych Can you give me more info about your cluster? I think tf-operator will not be the bottleneck.

[jtfogarty]

/area engprod
/priority p2

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.