ypserv loops after patching
I applied the latest (Aug/25/06) patch set the other day. On my NIS master ypserv, after a time greater then 10 minutes, goes into a loop taking over one out of my processors.
4979 root100 5576K 2176K cpu395:00 24.99% ypserv
In syslog there is a warning on startup.
ypserv[4979]: [ID 783678 daemon.warning] /usr/lib/netsvc/yp/ypserv: no /var/yp/securenets file
But securenets has never been required.
I don't know much about dtrace but from what I have extracted from it is:
errors:
1ypserv stat2 No such file or directory
1ypservfcntl22 Invalid argument
1ypservgetmsg11 Resource temporarily unavailable
1ypservopen642 No such file or directory
1ypserv lwp_wait45 Deadlock condition.
The deadlock interest me but I don't know how to figure out what is causin g the deadlock.
I tried a dtruss. It is mainly doing:
36254 pollsys(0x63A50, 0x5, 0x0)= 1 0
36261 ioctl(0x5, 0x530F, 0xFFBFFA3C)= 1 0
36263 fstat(0x5, 0xFFBFF960, 0xFEEC2000)= 0 0
36268 lwp_sigmask(0x3, 0x0, 0x0)= 0xFFFF 0
36269 lwp_sigmask(0x3, 0x2000, 0x0)= 0xFFFF 0
Not a huge help. The only thing it tried top open while I was watching it
was:
open("/dev/ticotsord\0", 0x2, 0x0)= 9 0
If someone could point me to a few other things to try to figure out why ypserv is gone into training for "America's Got Talent" the infinate loop edition, I would love to hear it.
Part of the process of figuring this out has been reading documentation, and for the first time in a while I've been very disapointed about the lack of man pages on Solaris. Did I forget to install a package or are there a lower standard for documentation coming out of Sun? For example I go looking for the man page for lwp_sigmask. It doesn't exist. sigmask does, but that syscall only takes 1 argument, when as we can see above lwp_sigmask takes 3.
I could back out of the patches, but I haven't had to do that under Solaris in ten years of administration and I would rather not start now.
Thanks for the help.
