You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If a network interface name is set and it doesn't exist, cctl will hang trying to create the Kubernetes cluster (actually ssh-provider will hang because nodeadm hangs waiting for kubelet to converge).
The text was updated successfully, but these errors were encountered:
Root cause is that keepalived is in an active but not operational state and kubeadm will wait for the control plane to be up using the API endpoint (vip).
Mon Jan 14 23:38:06 2019: Starting VRRP child process, pid=7
Mon Jan 14 23:38:06 2019: Registering Kernel netlink reflector
Mon Jan 14 23:38:06 2019: Registering Kernel netlink command channel
Mon Jan 14 23:38:06 2019: Opening file '/usr/local/etc/keepalived/keepalived.conf'.
Mon Jan 14 23:38:06 2019: WARNING - interface ens3 for vrrp_instance K8S_APISERVER doesn't exist
Mon Jan 14 23:38:06 2019: Non-existent interface specified in configuration
Mon Jan 14 23:38:06 2019: Stopped
Mon Jan 14 23:38:06 2019: VRRP child process(7) died: Respawning
Keepalived will not fail fast upon receiving an "invalid" configuration. (It is possible that the interface may come up later). A timeout for nodeadm init could prevent indefinite hangs in the future.
Add the interface check to nodeadm might be preferred, so as to not run into the same issue when running init manually on the node.
If a network interface name is set and it doesn't exist, cctl will hang trying to create the Kubernetes cluster (actually ssh-provider will hang because nodeadm hangs waiting for kubelet to converge).
The text was updated successfully, but these errors were encountered: