Tim Allclair
Tim Allclair
> I presume that the spirit of your proposal is doing something less intrusive or risky It's not only about reducing risk, it's also about minimizing complexity. It looks like...
I spoke with a few of the konnectivity maintainers today, and they agreed they're interested in further exploring this path, and also that that means a KEP update. Please tag...
I think that with the current way we use gRPC this wouldn't do anything for us. My understanding that this would affect connections to the gRPC server when there aren't...
/remove-lifecycle stale
/reopen /lifecycle frozen
One way this could be improved is with an option to exclude transitive dependencies. I think those edges lose their value in large graphs and add a lot of noise.
/triage accepted /priority important-longterm It looks like eviction has some creative uses of `TooManyRequests` errors. Why is this conflict error converted to TooManyRequests? https://github.com/kubernetes/kubernetes/blob/786316f0b6bb9078cd564ebf5401bb2e9ac7f2a2/pkg/registry/core/pod/storage/eviction.go#L294-L305 Also this observed generation mismatch (stale...
I agree with this direction. A lot of memory leaks we've seen have to do with the extra state transition between pending dial & connected. Using a consistent ID across...
I'm probably getting ahead of myself, but we might want to add the new `TunnelUID` field to the top-level `Packet` message, since every packet type should have it. That would...
##### Alternative proposal for migration plan **Summary:** Each component handles _incoming_ packets with only old identifiers (random/connectionID), only new identifier (tunnelUID), or both. Each component outputs packets that always include...