Open ibra-coul opened 1 month ago
@MITCOULIB I also noticed this but I think it is because the helm chart uses a very old KV-CP version by default, 0.0.4: https://github.com/kube-vip/helm-charts/blob/main/charts/kube-vip-cloud-provider/Chart.yaml#L23 Compare to https://github.com/kube-vip/kube-vip-cloud-provider/tags
When I added to my values:
image:
tag: v0.0.10
the newer version was deployed with much improved features. In particular, it read the configmap from the namespace configured by the env var. It also prints a new log message "provider.go:86] Watching configMap for pool config with name: 'kv-cm', namespace: 'kube-vip'" to confirm.
So I think this is just a side effect of old versions
@rptaylor I set the correct version to v0.0.10 in the chart and tested everything works fine. If the pr #48 is merged it will be up to date.
This can be closed thanks to https://github.com/kube-vip/helm-charts/pull/48
kube-vip-cloud-provider: when I define the environment variables KUBEVIP_NAMESPACE and KUBEVIP_CONFIG_MAP, kube-vip-cloud-provider creates the configmap in the correct namespace and also creates in kube-system another empty configmap with the same name. It is the empty confimap that it uses so it is impossible to distribute the IPs. If kube-vip-cloud-provider is deployed in the kube-system namespace with the KUBEVIP_NAMESPACE and KUBEVIP_CONFIG_MAP environment variables, it creates the configmap with the right name and everything works fine.