Hello,
My GKE cluster was auto-updated from Kubernetes 1.32 to 1.33 this weekend (I was on the rapid channel). After the update, the Envoy ingress pods (managed by Contour, injected with linkerd.io/inject : ingress ) started flip-flopping between Ready and not. Turns out the Linkerd proxy’s liveness probe (http://:4191/live ) was failing with context deadline exceeded (Client.Timeout exceeded while awaiting headers) .
Removing the Linkerd proxies from the Envoy pods (linkerd.io/inject : disabled ) fixed the issue.
I was still on Linkerd enterprise-2.15.4 , so I updated to 2.18.1 . I re-enabled proxy injection in Envoy, and at first it was working but after a few hours I started seeing the same behaviour again (liveness probe failing intermittently).
So I disabled proxy injection again and everything is working for now, but I’d like to get to the bottom of this and ideally be able to re-enable proxy injection in my Envoy pods.
According to this, the maximum supported Kubernetes version is 1.32. Does that mean that there are known issues with 1.33, or simply that it hasn’t been tested with 1.33?