chartmuseum
chartmuseum copied to clipboard
helm
Chart version missing but the blob file exists
{
error: "no chart version found for web-redesign-1.2.2"
}
However, the azure blob file does exist
This randomly happens to many other charts, we don't have any redis cache, and memory usage seems fine in our pods
kubectl top pod -n chartmuseum
NAME CPU(cores) MEMORY(bytes)
chartmuseum-64ccf6448b-cjk5q 66m 52Mi
chartmuseum-64ccf6448b-fbvgs 1m 46Mi
@tomqwu if you aren't using Redis as an external cache, each ChartMuseum pod maintains an in-memory cache of the index file. If you are running multiple ChartMuseum replicas, they will each maintain their own index and will be out of sync.
You may be able to leverage the --cache-interval to scan the backend storage for changes on an interval to keep the index up to date or switch to using Redis for caching.
Single pod, multiple pods, the behaviour is the same. We also tried to redeploy with the new pod. Regardless of the number of pods, all pods show missing chart versions.
The total blob files is less than 1000 binaries, and total size is less than 6mb
@tomqwu what version of ChartMuseum are you running? I’d recommend upgrading to the latest version (if you aren’t already)
if you want to sync whats in the backend storage with the index served by ChartMuseum, you can use the --cache-interval flag.
There was an issue in previous versions of ChartMuseum that could have caused this issue but upgrading to a new version may not retroactively "fix" the problem without using the cache-interval flag.
Once the index is up to date with whats in your backend storage and you are using the latest ChartMuseum, I'd be interested if you run into this again.
It’s interesting I thought about the cache interval, however we’ve made a few new deployment too without any previous state with same result, since our storage option is blob, there was no statefulset used.
On Oct 28, 2022, at 2:47 PM, Casey Buto @.***> wrote:
if you want to sync whats in the backend storage with the index served by ChartMuseum, you can use the --cache-interval flag.
There was an issue in previous versions of ChartMuseum that could have caused this issue but upgrading to a new version may not retroactively "fix" the problem without using the cache-interval flag.
Once the index is up to date with whats in your backend storage and you are using the latest ChartMuseum, I'd be interested if you run into this again.
— Reply to this email directly, view it on GitHub https://github.com/helm/chartmuseum/issues/635#issuecomment-1295339837, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABGK3ITULGRX3H5G7QYVXD3WFQNV5ANCNFSM6AAAAAARLISA3U. You are receiving this because you were mentioned.