Hello - I’m trying to understand how to configure destination rule http2maxrequests settings in conjunction with Auto scaling of the pods comprising the callers (mostly the ingress gateway) and the pods comprising the service being called.
Based on reading and on experimental testing completed myself, it appears that the http2maxrequests field controls 2 different behaviours.
a source envoy - eg an ingress gateway instance - will make at most ‘http2maxrequests’ concurrent requests to the virtual service, regardless of the # pods serving the virtual service
the destination envoy - eg the sidecar on the service pod - will forward at most ‘http2maxrequests’ concurrent requests to the service container in the pod.
Do I understand the behaviour correctly? If so, are there any best practice guides for configuring this with autoscaling in place for both client and destination sides of the flows?