Closed felipecrs closed 7 months ago
@felipecrs thanks for creating this issue! It looks to me you are using a custom helm chart that still includes the install hook we had in version v0.15.7, that was removed in 0.17, which is probably why this fails, because there shouldn't be any post-install hooks with the latest versions anymore.
@FabianKramm, that's weird. I'm not using any custom helm chart, I'm just using --distro k8s and --kubernetes-version 1.27. But I'll do some investigation here.
Oh, I see the confusion here. @FabianKramm, the install-hook mentioned in my logs are actually the logs of when I try to install my application against the vcluster.
I posted the vcluster logs after those logs.
@felipecrs that might be a timing issue I believe, vCluster might take a little longer to startup as some replicas might not be up yet or are restarting, we will refactor the k8s distro for 0.19 to include similar to how we do it with k3s and k0s the binaries in the syncer container which should also solve this problem.
Got it. Will wait for 0.19 then. Thanks a lot.
Confirmed to work with vcluster 0.19.0. Thanks a ton.
Ok, to be fair I realized it only worked because I had upgraded other stuff in my services. If I try vcluster 0.19.0 on an old version of my services, it still fails with the same problem (but with vcluster 0.15.7 works fine).
However, this is no longer a problem for me. Just posting in case someone finds the same issue and reads this.
What happened?
I have a CI/CD pipeline that creates a vcluster and deploys some workload to it. When I was using vcluster 0.15.7, everything worked fine. After upgrading to latest version, 0.18.1, it's no longer working. I also noticed the exact same issue at least with 0.17.1 and 0.18.0 as well. Possibly also 0.16.0.
Note that this issue is a little different from https://github.com/loft-sh/vcluster/issues/1285. There, most of my workloads stopped working after upgrading from 0.15.7 to 0.16.0, but started working again after 0.17.1.
This time, a specific workload/pipeline is not working.
This is the apparent failure in the pipeline, which happens during a given helm install we have:
Listing the vcluster pods:
Checking logs for the
vcluster-api-748f6596f-88qgc
:What did you expect to happen?
To work as it worked with 0.15.7.
How can we reproduce it (as minimally and precisely as possible)?
Sorry, I was not able to create a minimal reproduction environment, but I can easily and reliably replicate the error here. I can also replicate the fact that 0.15.7 works.
Anything else we need to know?
Here is how I create the vcluster:
Host cluster Kubernetes version
Host cluster Kubernetes distribution
vlcuster version
Vcluster Kubernetes distribution(k3s(default)), k8s, k0s)
OS and Arch