Here is ktrace output for one I suspect to be dead. Unfortunately having ktrace log to a NFS mount causes panics so the box rebooted and I lost all the dead processes I was saving up to test on :-) All this process is doing is the following over, and over, and over... 960 nntpcached PSIG SIGALRM caught handler=0x2005e660 mask=0x2000 code=0x0 960 nntpcached RET sigsuspend -1 errno 4 Interrupted system call 960 nntpcached CALL sigreturn(0xefbf355c) 960 nntpcached RET sigreturn JUSTRETURN 960 nntpcached CALL sigaction(0xe,0xefbf35c4,0) 960 nntpcached RET sigaction 0 960 nntpcached CALL sigprocmask(0x3,0) 960 nntpcached RET sigprocmask 8192/0x2000 960 nntpcached CALL setitimer(0,0xefbf35dc,0) 960 nntpcached RET setitimer 0 960 nntpcached CALL setitimer(0,0xefbf35ec,0xefbf35dc) 960 nntpcached RET setitimer 0 960 nntpcached CALL sigaction(0xe,0xefbf35d0,0xefbf35c4) 960 nntpcached RET sigaction 0 960 nntpcached CALL sigprocmask(0x1,0x2000) 960 nntpcached RET sigprocmask 0 960 nntpcached CALL setitimer(0,0xefbf35ec,0) 960 nntpcached RET setitimer 0 960 nntpcached CALL sigsuspend(0) It looks like it is getting the alarm() due to client timeout, but not properly handling it. Evan -----Original Message----- From: proff@suburbia.net <proff@suburbia.net> To: Evan Champion <evanc@synapse.net> Cc: nntpcache-users@suburbia.net <nntpcache-users@suburbia.net> Date: Wednesday, April 08, 1998 5:25 AM Subject: Re: NNTPC: Problem with dead nntpcache sessions? >> >> >What I've noticed is that if there's a reboot while a hung session >> >is open, there's a fairly good chance that cache.mmap will be >> >corrupted, preventing nntpcached restarting (and generating >> >shitloads of errors if it's being run out of inittab) >> >> >> I've been using anonymous mmap()'s so that hasn't been an issue for me. >> However, every couple days I go cleanup the nntpcache processes and I >> usually end up terminating around 20, so that is a pretty significant dead >> server count. It is especially bad if those connections are holding open >> NNTP sessions to outside servers, because I'm being charged based on my >> maximum concurrent NNTP session count :-) [note: I haven't actually checked >> to see if this is the case]. >> >> Evan > >Can you ktrace -p or strace -p oen of these processes to see what it's doing? >What does the proctile info say? > >Cheers, >Julian. >
