Bind is hanging on CentOS 4.4

JINMEI Tatuya / 神明達哉 jinmei at isc.org
Fri May 29 21:56:51 UTC 2009


At Fri, 29 May 2009 15:17:23 -0400,
"Jesse Cabral" <jcabral at mtsolutions.net> wrote:

> So I can understand the original goal, let me re-clarify the objective.
> 
> The problem of Bind hanging is thought to be caused by an interthread lock.

First off, it's just one possibility.  "Hanging" is a pretty general
symptom so we cannot be sure if it's really a thread problem.

> Let me ask this questions, is the goal to disable threads on
> multi-processors or threads completely ?

When I suggested it I meant the latter.  In fact, if a threaded
version only uses a single core or processor thread-related, many of
the thread related bugs such as race condition or deadlock are
unlikely to happen.  But in a process of identifying a cause of the
trouble, I'd suggest starting with a solid base, which, in this case,
is completely disabling threads.

Besides, the following strongly indicates you are somehow confused in
your debugging:

> ps -Leo user,pid,ppid,lstart,lwp,nlwp,psr,args |egrep "LWP|named"
> USER       PID  PPID                  STARTED   LWP NLWP PSR COMMAND
> named    17829     1 Fri May 29 14:57:46 2009 17829    4   1 /usr/sbin/named
> -u named -n 1 -t /var/named/chroot
> named    17829     1 Fri May 29 14:57:46 2009 17830    4   1 /usr/sbin/named
> -u named -n 1 -t /var/named/chroot
> named    17829     1 Fri May 29 14:57:46 2009 17831    4   1 /usr/sbin/named
> -u named -n 1 -t /var/named/chroot
> named    17829     1 Fri May 29 14:57:46 2009 17832    4   1 /usr/sbin/named
> -u named -n 1 -t /var/named/chroot

that is, you thought you disabled threads but still saw multiple
threads in running named.  Whether you go without threads altogether
or with threads on a single processor, you should first be sure about
what you're doing.

---
JINMEI, Tatuya
Internet Systems Consortium, Inc.



More information about the bind-users mailing list