when cluster is running, i input shell kill or stop smbd on one node, have a moment the ctdb display this node unthealy, and make recovery. so i have a question: why not ctdb try to restart smbd?
I also wonder about it.
This is by design. Samba process should never go away once it's started from CTDB startup event. If Samba process goes away, then either it has crashed or someone has terminated it. In both cases, it's important to know what has happened and why. That's why CTDB will never try to restart Samba processes.
This was explained a long time ago and there have been no further questions. Closing with reason "won't fix", since current behaviour is as per design.