Quoth Keith Cantrell on Wed, Oct 22, 2008 at 02:57:08PM -0500: > This is indiciative of a kernel bug -- processes should die when > SIGKILLed. > > Agreed, but arguably, svc.startd shouldn't try forever, meanwhile > filling up the filesystem with ever increasing log files.
Sure. File an RFE. > Yes, init is still running and no 'init' messages in the messages file > or dmesg. What is pointing you to init? I guess because it can't > seem to reap the process? I'm guessing the offending process is stuck > in kernel land and refuses to come out, which is why 'init' can't get > rid of it. I still think svc.startd should have some kind of fail > safe that if a process wont die after a few hundred attempts, it > should give up. Yes, except if it were still active and just stuck in a system call, I wouldn't expect ps to display it as "<defunct>". I don't know a lot about the process model, though. David
