John Howard
John Howard
I think keepalives are a good general practice. There is no guarantee a TCP connection stays live -- there can ALWAYS be something that non-gracefully terminates. But they also shouldn't...
I don't think we should let these decisions tie to whether we use keepalives. They are either useful or not to detect non-graceful broken connections, and that doesn't really change...
One thing I am concerned with (1) is, consider a case where I have `pod-a` sending traffic to `pod-b`, even after termination has started: With (1), once pod-b starts to...
> Right - in both cases, all the connections the user app (pod-a) is making will fail (from their perspective) No, that is not the case. Maybe I am not...
One thing to clarify is GOAWAY doesn't close the inner connections. It just means once there no more active inner connections, the out connection will be dropped instead of saved...
I think the issue is probably that NONE mode cannot work on a gateway. NONE relies on original destination, which only applies to sidecars.
I think the only thing left is text/template. blocked by https://github.com/golang/go/issues/72895 probably
Grafana made a change at some point that made linking to a dashboard by name impossible. The problem is our dashboards do not know the GUID of the dashboards other...
Not certain but I think as part of https://github.com/istio/istio/issues/50127 we could assign the dashboards a static UID that we know about
No, still need work