Bug: The dsChecksum wasn't updated on the autoscaling request

[JKBGIT1]

Current Behaviour

On the autoscaling request Claudie didn't change the dsChecksum, only the desiredState. Due to that, the autoscaler didn't work (see).

According to the logs, the cluster-autoscaler failed to retrieve the resource lock kube-system/cluster-autoscaler and then lost master. The same error occurrence #1065 (comment) .

$ kubectl logs -n claudie autoscaler-wox01-cluster-qy5w5zl-57f74c7dd5-v8gbd -c cluster-autoscaler -p
...
I0417 18:08:17.383459       1 static_autoscaler.go:673] Decreasing size of compute01-ccx23-auto-fy7ww3o, expected=7 current=6 delta=-1
I0417 18:08:17.383801       1 static_autoscaler.go:426] Some node group target size was fixed, skipping the iteration
I0417 18:08:27.485122       1 static_autoscaler.go:673] Decreasing size of compute01-ccx23-auto-fy7ww3o, expected=7 current=6 delta=-1
I0417 18:08:27.485366       1 static_autoscaler.go:426] Some node group target size was fixed, skipping the iteration
I0417 18:08:28.291313       1 node_instances_cache.go:156] Start refreshing cloud provider node instances cache
I0417 18:08:28.292135       1 node_instances_cache.go:168] Refresh cloud provider node instances cache finished, refresh took 787.249µs
E0417 18:08:33.691966       1 leaderelection.go:330] error retrieving resource lock kube-system/cluster-autoscaler: Get "https://loadbalancer.worldofpotter.eu:6443/apis/coordination.k8s.io/v1/namespaces/kube-system/leases/cluster-autoscaler": context deadline exceeded
I0417 18:08:33.695145       1 leaderelection.go:283] failed to renew lease kube-system/cluster-autoscaler: timed out waiting for the condition
F0417 18:08:40.377907       1 main.go:578] lost master

Besides that, there was an error in the builder from 8 days ago. The error was produced by the ansibler due to the timeout on static nodes while installing the VPN. However, the InputManifest was in a DONE state constantly... (see #1339)

One more thing. The cluster-autoscaler states that there are 6 nodes and the expected amount is 7. On the other hand, the InputManifest record in Mongo has 7 autoscaled nodes in the currentState and 6 autoscaled nodes in the desiredState.

Expected Behaviour

Claudie should update the value of the dsChecksum when it updates the desiredState.

Steps To Reproduce

I don't know.

Anything else to note

The same error in the cluster-autoscaler #1065 (comment)

EDIT: a workaround for this error, when it appears.