The default time for Server Stop is 30 minutes. You can change that with the tsm stop --request-timeout <timeout in seconds> and use say ten minutes. Try that. I suspect it's just taking awhile to stop processes that are busy. They should not be on a TEST node since no user would be using that Server node. Do you think users are connected to that Server other than a stray administrator?
I'm also curious why you would have multiple nodes on a TEST or DEV server. They are not PROD and normally customers would not include other Nodes on DEV or TEST.
Our building new test cluster that should be replica of our PROD. Right now no user has access except my self.
whenever we tried tsm restart and tsm stop, it showing server is running, but node 2 n node 3 are stopped at that time.
You might want to log a ticket with support to have them determine why the ism commands are not completely stopping or restarting your main server. Something odd going on. It could be some type of time-out issue and they can recommend changes.