Octave Orgeron wrote:
> Have you tried preap against the process? That's how I get rid of zombie
> processes and things that just don't respond to kill properly.
Nah, preap won't work:
[EMAIL PROTECTED] ~]# ps -ef | grep 15540
webservd 15540 1 0 Nov 06 ? 0:09 /opt/local/sbin/httpd
-k start
[EMAIL PROTECTED] ~]# kill -9 15540
[EMAIL PROTECTED] ~]# ps -ef | grep 15540
webservd 15540 1 0 Nov 06 ? 0:09 /opt/local/sbin/httpd
-k start
[EMAIL PROTECTED] ~]# pstack 15540
pstack: cannot examine 15540: unanticipated system error
[EMAIL PROTECTED] ~]# preap 15540
preap: process not defunct: 15540
[EMAIL PROTECTED] ~]# truss -p 15540
truss: unanticipated system error: 15540
[EMAIL PROTECTED] ~]# dtrace -p 15540 -P syscall {}
dtrace: failed to grab pid 15540: unanticipated system error
I went through this on the list once, I think the solution we came to
was, after a ton of tweeking in mdb, forcing it to zombie and then
reaping it.... but that only worked some of the time. In any case,
doing surgery on a hung process via mdb on a running system is
potentially more dangerous than just rebooting the system, even if its
serving dozens of customers. A 3 minute reboot is far more preferable
to a system panic and ensuing cleanup.
benr.
_______________________________________________
sysadmin-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss
