Why nodeagent can't be started from Dmgr?

The dmgr uses the nodeagent process to execute commands on a remote machine. In theory, an alternative would be to use ssh to execute remote commands, but WebSphere Application Server does not use SSH because it would add too much extra complexity (SSH keys, OS user ID management, not easily supported on all platforms supported by WebSphere, etc.). Consequently, when the nodeagent is down, there is no way for the dmgr to execute a startNode.sh/.bat command on the remote machine to start it. In theory, the dmgr could execute the command itself if the node is on the same machine, but IBM doesn't recommend putting the dmgr and other nodes on the same machine anyway, so there's no reason to add that capability
In distributed server configurations, each node has a node agent that works with
the deployment manager to manage administration processes. A node agent is created automatically when you add (federate) a stand-alone node to a cell.

To add few more points, if you understand how any application server works then you will understand why the nodeagent in WAS or node manager in WLS cannot be should not be started remotly. Basically they are admin agents so if you allow to start them using a remote script then that itself a big security hole. So triggering them remotely using any any script is not possible. To start or any activity to happen in a remote server you need a process to listen in HTTP or any communication protocol which can authenticate the request and process it. But any of this is not p


Possible when NA or NM is down.

If NA will be down - DMGR wont be able to communicate to that node which in turn take us back to 1st step --- DMGR wont be having any info of that node -> if you dont know whom to contact then how will you give him/her instructions -- thats why we have restart and stop option in DMGR and no start option.