Feature request: make the action taken by skypilot-status-refresh-daemon user aware

[aylei]

I launched a cluster on my local Kind and the cluster is cleaned after I restarted my docker daemon:

$ sky status
Clusters
NAME                          LAUNCHED    RESOURCES                         STATUS   AUTOSTOP  COMMAND
aylei3                        1 min ago   1x Kubernetes(1CPU--1GB, cpus=1)  UP       -         sky launch -c aylei3 --cpus...

kg po
NAME                   READY   STATUS    RESTARTS   AGE
aylei3-57339c81-head   0/1     Unknown   0          15m

Of course the Pod is down since the kind node (runs as container in docker daemon VM) is restarted. The sky cluster aylei3 also disappeared after a while. It turns out there is a daemon that refresh the cluster status every 60s, the daemon log explains why:

$ sky api logs skypilot-status-refresh-daemon
....
Cluster terminated on the cloud: aylei3
Failed to refresh status for 1 cluster:
  test: 'test' (Kubernetes) is owned by account ['gke_sky-dev-465_us-central1-c_skypilotalpha_gke_sky-dev-465_us-central1-c_skypilotalpha_skypilot'], but available identities are [['kind-skypilot_kind-skypilot_default'], ['kind-kind_kind-kind_default'], ['kind-skypilot_kind-skypilot_default'], ['temp_temp_default']]. Check your kubeconfig file and make sure the correct context is available.
I 02-27 19:07:50 requests.py:311] Status refreshed. Sleeping 60 seconds for the next refresh...
I 02-27 19:07:50 requests.py:311]
I 02-27 19:08:50 requests.py:306] === Refreshing cluster status ===

It is really confusing since my cluster is cleaned silently. It would be helpful if we:

• notify users that some actions have been taken by the refresh daemon in the next sky CLI invocation
• or (better), prompt user that the daemon discovers that some actions should be made, and ask user's confirmation to perform these actions

[concretevitamin]

+1. Before apiserver there's a hint printed on next command. Now with the refresh daemon it's hard to notice.