OK, just so we can be 100% sure.
If you find a job that restarts in the GUI, does this same job restart on the command line.
Likewise, for a job that fails to restart on the command line, does this same job ID restart in the GUI.
Just need to be vey sure you are testing the same jobid in both (GUI/ command) line cases.
I think from my post a few up, you did confirm this, I just want to be very sure.
Next steps.
Enable the bpjobd log at VERBOSE = 5 and restart a job that shows the issue and collect the log.
I can kinda reproduce this :
root@womble admin $ bpdbjobs -restart 970
Restarting 0 jobs
... but that is using a jobid that doesn't exist - that is I haven't got to jobid 970 yet.
I was wondering if the system canot find the jobid (even though it is there) but that makes no sense as you explained that the GUI restarted the job.
Oh, before I forget, as well as the admin log, will need nbpem
vxlogcfg -a -p 51216 -o 116 -s DebugLevel=6 -s DiagnosticLevel=6
That sets the log, use values of 1 or 0 to turn back down (DiagnosticLevel = 6 is default)
The log in raw format, is in
/usr/openv/logs/nbpem
Only get the log covering the time period the command was run.
Forget Java (nbjlogs) - not going to help in this case ... it doesn't log what I was hoping it would.