Docs - google_container_cluster - initial_node_count 1 for remove_default_node_pool = true causes auto repair

[jimsnab]

This doc page recommends

  # We can't create a cluster with no node pool defined, but we want to only use
  # separately managed node pools. So we create the smallest possible default
  # node pool and immediately delete it.
  remove_default_node_pool = true
  initial_node_count       = 1

Unless the separately managed node pool only has one node, GKE will go into repair mode after creation. GKE will spin for about 30 mins in repair mode - although the cluster is mostly usable.

The docs should recommend sizing the default node pool the same as the replacement node pool.

I only have one replacement node pool, so someone else will have to check the behavior when using two or more replacement node pools.

References

• b/298514262

[rileykarson]

We should document that the default node pool should be sized around where the separately managed pools will end up. That will reduce the chances (but not eliminate them) of a master resize.