Closed uptickmetachu closed 4 months ago
Hi @uptickmetachu, there are a few known issues with our 1.0 release with the kubernetes support, which is why we never documented it, it simply wasn't ready yet. We've completely reimplemented it in beyla:main and we believe that works well now. We still haven't released 1.1, but it should happen soon.
If you'd like you can try our main branch, we have examples in our tests with k8s support, as well as choosing what to instrument by pod name. But it's still an unreleased version.
Sorry this is crashing with 1.1.0.
The image is the one below.
Image: grafana/beyla:1.1.0
Image ID: docker.io/grafana/beyla@sha256:ae7997624fd99b250a1a533be34d339fde3328a24688c373dc4b2d3862941f4b
I also tried 1.0.0 and 1.0.2 and that fails on:
E0110 21:47:11.489684 19250 reflector.go:148] k8s.io/client-go/informers/factory.go:150: Failed to watch *v1.Service: unable to sync list result: couldn't enqueue object: not indexing service without ClusterIP
despite having cluster permissions to watch services.
I see, do you mind trying with grafana/beyla:main
please? I believe we cut some early pre-release 1.1 before the kubernetes work was finished.
Based on what I see in the error it may not help, but I just want to make sure we are not working on something that's already fixed. Thanks a ton in advance!
@mariomac this looks like a bug in the k8s library we use, or simply they may not support EKS?
panic: interface conversion: interface {} is *kube.PodInfo, not *v1.Pod
I see, do you mind trying with grafana/beyla:main please? I believe we cut some early pre-release 1.1 before the kubernetes work was finished.
That did the trick :). I'll pin it for now and continue playing with it. I'm eagerly awaiting the proper 1.1 release!
Just FYI, we just released 1.2 which has these changes, so you can pin to that, rather than main, which might bring surprises :).
Running beyla:latest (1.02?) as a daemonset on EKS (1.25) I get this error when I enable kube metadata parsing.