Julian Taylor

POC patchfor explicit nodeport check ( using NODE_IP environment variable ) ```patch --- a/speaker/bgp_controller.go +++ b/speaker/bgp_controller.go @@ -22,6 +22,7 @@ import ( "sort" "strconv" "time" + "os" "go.universe.tf/metallb/internal/bgp" "go.universe.tf/metallb/internal/config" @@...

I do not see how that kubernetes issue is related to this issue. This problem cannot be solved by kubernetes alone, it requires metallb consumes feedback from kubernetes when the...

> why metallb should be faster than kube-proxy? metallb announces the route when the endpoint is ready, kube-proxy configures the node network when the endpoint is ready metallb just does...

What did happen in kubernetes 1.28 is that the kube-proxy iptables implementation was significantly speed up for incremental updates. This is probably why some cannot reproduce this anymore. But even...

ok checking some more, related to upstreams question to metallb: > Your loadbalancer controller has to respect kube-proxy healthcheck to avoid the race, he can not start sending traffic to...

@diefans which version of vim? 7.4 has a new "faster" regex engine which is absolutely horrible for easytags in its current form (yes it allocates millions of memory blocks per...

see https://github.com/xolox/vim-easytags/issues/68#issuecomment-28480981