Failed to reconcile after changing ClusterIP to LoadBalancer

[markandrus]

We received the following error after changing our replica-service-template to use a LoadBalancer instead of ClusterIP:

-                type: 'ClusterIP',
-                clusterIP: 'None'
+                type: 'LoadBalancer'

Create Service clickhouse-v2/chi-v2-v2-0-0 failed with error Service "chi-v2-v2-0-0" is invalid: spec.clusterIPs[0]: Invalid value: "None": may not be set to 'None' for LoadBalancer services

Not sure how to reconcile. Maybe get rid of the replica-service-template and add it back?

[alex-zaitsev]

@markandrus , could you try to remove clusterIP?

[markandrus]

That worked: I deleted the service which was assigned the ClusterIP before redeploying my changes. 👍 Thanks

[markandrus]

I guess it's atypical to do this on an existing cluster? We are probably going to leave the ClusterIPs for internal replication and move to managing LoadBalancers ourselves. Our LoadBalancers are only for ingestion, and we don't way to pay the overhead of replicating through.

[alex-zaitsev]

Fixed in 0.23.x. This is actually a Kubernetes bug, so service needs to re-created when type is changed.