We are using Istio 1.0.6 and Istio 1.1.14 on a Production environment with Kubernetes 1.10.
We have approximately 100 services and run with more than 200 Pods.
Last week, we’ve experienced a lot of CPU peaks from Istio Pilot Discovery service. It usually consumes up to 1000m CPU, but sometimes it gets to 4000m and higher.
I’ve already configured pilot agent concurrency to 1, as suggest in a GitHub issue and also limited CPU to 3000m on pilot deployments, but we still get some CPU peaks.
Do you guys have any thought on that?
Thanks in advance!