Usually OKD offers the possibility to launch a command shell to debug a running
application. Since a pod that is failing may not be started and not accessible
to rsh or exec, the debug command makes it easy to create a carbon copy of
that setup. This can also be applied to nodes by running a debug pod.
The debug pod is deleted when the the remote command completes or the user interrupts the shell.
To run a script in an automated fashion on all nodes in the cluster without using SSH you could run:
[okd@services ~]$ sudo nodes=$(oc get nodes -o name)
[okd@services ~]$ sudo read -r -d '' commands <<- EOM || :
hostname
sudo nmcli con s
EOM
[okd@services ~]$ sudo for node in $nodes; do oc debug --quiet=true $node -- chroot /host /bin/bash -c 'tmp=$(mktemp) && echo "$0" > $tmp && . $tmp' "$(echo "$commands")" doneUsing SSH is not recommended to access a node and should be used with precaution
and only if necessary. Usually this is done when nodes do not properly boot and
running oc debug node/... becomes impossible as the pod is unschedulable when
the node is not ready.
[okd@services ~]$ sudo ssh core@$NODE -i ~/.ssh/fcos