[Asterisk-Users] Voicemail with NFS

[Tzafrir Cohen]

On Fri, Jun 16, 2006 at 02:17:53PM -0600, Douglas Garstang wrote:
> I have /var/spool/asterisk/voicemail NFS mounted from another server. 
> Everything is fine, until I simulate an NFS server failure, by 
> shutting down the NFS server process.
> 
> At this point, Asterisk becomes almost non-responsive. It won't even 
> process a 'sip show peers' command correctly. It displays a few lines 
> of text, pauses for several seconds, and then displays the rest. When 
> a call comes into the system, Asterisk seems to do nothing for several 
> seconds, and generally acts really sluggish. The phone gives up after 
> several seconds, because Asterisk isn't doing anything.

Why is Asterisk hung?

Is it just a long wait, or are some threads in a D state
(uninterruptable sleep)?

Maybe one thread is hung and is holding a certain lock and thus all
others are hung.

On kernel 2.6, ps won't show the different threads of the process by
default. Try:

  ps -T PID_OF_ASTERISK

Are different threads in "STAT" (state) D or S?
You can get a stack trace from at least from threads that are not in D
state using gdb:

  gdb -p PID_OF_THREAD
  # in the (gdb) prompt:
  backtrace
 
Though IIRC, the thread that is responsible to answer to 'sip show
peers' is actually the "main" thread (whose PID is the same as the "PID
of asterisk").

> 
> I have used the soft option with the NFS mount.

BTW: what version of NFS? Have you verified that the "soft" option is
indeed used? (take a look at /proc/mounts ).

BTW: the fact that you use Outlook makes it much more difficult to read
complex (branching) list threads in which you participate. I had to
spend some 15 minutes or so just to glue sub-threads together. And I'm
not sure most mailers even have that option to manually attach/detach
threads.

-- 
Tzafrir Cohen      sip:tzafrir at local.xorcom.com
icq#16849755       iax:tzafrir at local.xorcom.com
+972-50-7952406           
tzafrir.cohen at xorcom.com  http://www.xorcom.com