John Howard
John Howard
@hzxuzhonghu per https://github.com/istio/istio/issues/47593#issuecomment-1912384696 I don't think we can do this, we would need to create a webhook instead. The webhook is somewhat problematic since there is a circular dependency (webhook...
Why is it not long enough? It is configurable: terminationDrainDuration
> after receiving SIGTERM, envoy begin to drain listeners, it can still accept new connection but it send go away. +1. Although I think after 45s it will stop them...
The prestop probably won't do anything. Immediately after it will trigger the old behavior with shutdown after 45s. We probably need to change Istio's handling to always use skip_exit. I...
You will hit: > // If we drained now, skip draining + waiting later > // When we terminate, we will instead exit immediately immediately after preStop hook, kubelet will...
I actually tried the existing state without skip_exit and it seemed like even after 45s it never exited. Not sure if I did something wrong On Wed, Dec 6, 2023...
> Would you agree that MINIMUM_DRAIN_DURATION would be the solution instead of terminationDrainDuration in this case? > > Maybe this is just a documentation request? :+1: :+1:
I want to do some more investigating usage before merging. will report back soon
also the point of the experimental command was to replace the old one. This does the opposite. What changed?
`x ps` has ~3 modes: 1. Query each istiod and portforward 2. xds-via-agents 3. xds-address (1) just changes from HTTP to grpc/xds, which is not a concern and seems like...