steeling

Adding an excerpt from a comment above: Simply put, we shouldn't block on any routine that reads from a channel. This alleviates 4 instances of such blocking: 1. RateLimitedAdd from...

> We put a large object onto the queue, and we should really be putting just a key value for the object so we get the advantage that the [docs...

> > There's a lot of requests to provide performance numbers. I get the desire to do that, however we unfortunately lack good tooling to test this. I'll try to...

> > We have proof that the rate limiter is detrimental to our memory performance, > > The way we are using the rate limiter workqueue is causing issues as...

> > Note that it's only the proxy update pubsub that is coalesced, while the kube events pubsub is not, so no changes for those uses > > the events...

> Hey, can you modify https://github.com/openservicemesh/osm/blob/main/.github/workflows/auto-assign.yml#L3 to react on `pull_request_target` instead of `pull_request`? Looks like a weird permission issue is preventing the auto-assign done, let's see if that fixes it

Some useful resources to get started: https://www.envoyproxy.io/docs/envoy/latest/api-v3/extensions/filters/http/ext_authz/v3/ext_authz.proto#envoy-v3-api-msg-extensions-filters-http-ext-authz-v3-authorizationrequest https://github.com/salrashid123/envoy_external_authz/blob/master/authz_server/grpc_server.go

The `PortForwardToProxy` method should exist on the compute client in `pkg/compute/kube`

The `PortForwardToProxy` method should exist on the compute client in `pkg/compute/kube`

I'd personally prefer to see a PR like this come before https://github.com/openservicemesh/osm/pull/5073. I think there's a lot of caveats like how many can run at once (might infer if we...