"can't reach node 'master'" is a new one for us I'm afraid. I'd recommend simply killing the icebox and icegridnode processes with SIGTERM (give them some seconds before resorting to SIGKILL) then once you are sure the processes have gone then removing the .lock files from inside your binary repository, e.g., probably those reported by, find /OMERO -name .lock
I think you should then expect to be able to start up the server processes just fine. Still, weirdness can happen when one runs out of space, so the cautious may wish to back up before restarting the server, and if you do see any oddness, or require any elaboration of the above, then please don't hesitate to ask.
Cheers, and good luck,