[asterisk-dev] possible zaptel problem with SMP and RAID1

[critch]

On Mon, 2007-07-09 at 18:23 +0200, François Delawarde wrote:
> Hi,
> 
> I believe that it's a zaptel related problem, as I have the same 
> problems trying with OpenVox hardware (A400 and OPVXA1200) and Digium 
> hardware (I could test a TDM400P for a while, but I'm not a lucky owner 
> yet), so it appears to be hardware independent. We also have same 
> results on two standard motherboards, two different SMP CPU, different 
> hard drives in RAID, ...
> 
> I contacted OpenVox, they told us that it should be a software related 
> problem. Also contacted Octasis, as the problem occured with their echo 
> canceller and we first thought it was the reason, but they proved us 
> wrong (telling that the "bips" were inserted before the echo canceller, 
> maybe by zaptel modules). I then tested without echo canceler with same 
> results.

Your problem is "software" related. The zaptel cards require you
actually get back round to read and handle the audio data regularly. You
need to be able to let asterisk run in realtime so it can move these
frames around.

Your problem is going to be related to sysadmin work. You shouldn't run
software raid on the machine doing asterisk. RAID1 just makes your life
worse, you are being required by kernel level code to write everything
to disk multiple times. Once for every disk in the RAID1. My guess is
you have a SATA1 drive set, or possibly a poor SATA controller. When
combined with software RAID you are getting hung up in some kernel code
somewhere and by the time asterisk can service the channel again you
have lost audio.

Your test should be to write to the disk while you are listening to the
channel, or recording it. Tuning the drive subsystem will reduce your
problems. Removing software raid will further reduce your problems. A
good overall system pruning and tweaking is in order.
-- 
critch <critch at basesys.com>