kibbles-n-bytes
commented
3 years ago Hey @Scorpiion , creating a new GKE cluster on the regular channel will today install version 1.23.0 of Config Connector. Add-on updates are triggered by every master node upgrade, which on the regular channel should have certainly happened between when it looks like the cluster was installed and now.
Could you share the gcloud container clusters describe output for that particular cluster? And in addition, could you share the output of the following call:
kubectl get po -n configconnector-operator-system configconnector-operator-0 -o jsonpath='{.metadata.annotations.cnrm\.cloud\.google\.com/operator-version}'We do unfortunately have a time lag between our standalone release and our add-on release on the order of ~3 weeks, which we are looking to reduce in the future. However, the lag you're describing is way too long, so this must be some sort of issue either with your cluster configuration or internally on the Google API.
You described two problems in this thread, though; the title is about the lag being too long, but the VM problem seems separate. Could you confirm on a separate cluster on 1.27.0 if the issue still remains? Or, let's wait until we understand your add-on upgrade issue, and then see if it is still an issue at that point in time.
Scorpiion
toumorokoshi
Describe the bug We have two problems with config connector right now:
The VM got setup correctly when the yaml was initially applied, but then config connector started giving these errors and now we are in a state where we can't do updates and this is a quite big blocker right now (some things I have manually updated in the UI and "backported" to the yaml, or the other way around, but it's an accident waiting to happen). I have had this for about a month now, thinking that a GKE addon update would come soon that might resolve this but so far no luck. Initially we used the "stable" GKE release channel, but I updated us to use the "regular" channel so we should be able to use the GKE addon (previously had a script based automation of config connector install/update). Now we seem to be "stuck" with config connector version
1.15.1that was released in 2020-03-19 (https://github.com/GoogleCloudPlatform/k8s-config-connector/releases/tag/1.5.1), that is more than 7 months ago. And to be honest, with the high development pace of config connector that is ages...One of our problems above with
networkInterface.0.NetworkIpI think is resolved in 1.6.1 as described here. But yeah, even if it was fixed in April I don't seem to be able to get this update with the addon.The second error above about the bootdisk, it seemed when I tested earlier that only happened when using the
cos-cloud/cos-81-ltsvalue asbootDisk.initializeParams.sourceImageRef.external, when I used the debian family of image the error did not appear I think (but it was some time ago I tested, I'm 95% it did not appear with debian). So I've been thinking that this maybe has also been resolved already in newer versions of config connector but not completely sure.ConfigConnector Version
To Reproduce Install GKE with regular channel and config connector addon. Check that version is 1.15.1. I don't think reproducing the error I got make sense time wise at the moment, I think it's more important to update the config connector addon so I can check with a later version.
Questions
cnrm.cloud.google.com/deletion-policy: abandon) and then recreate it but it gives the same error and no updates of for example the metadatauser-datavalue that I want to update.If I have missed some information above please just ask and I will try to provide it ASAP.