We are using an Asterisk box to route VOIP calls through to a traditional
telephony IVR. At low to moderate call volumes (say < 20 calls simultaneous) it
seems fine. Once the call load approaches capacity we get HDLC Abort and HDLC
Bad FCS errors and some calls in progress get badly distorted audio, to the
extent that a conversation would be unintelligible.
Current Setup:
IBM x3650 M2
Fedora 12
Asterisk 1.6.2.8
OpenVox D410P
E1 crossover cable to a Dialogic based IVR, which has another trunk connected to
the PSTN which does not experience problems.
1. Prevent X Window GUI from starting
2. Swapped out E1 crossover cable
3. Changed OpenVox D410P DIP switch 7.2 (PCI burst setting)
4. Changed Dialogic timing service on separate IVR to take timing for this trunk
from the PSTN trunk.
5. Tried using the HDLC hardware setting rather than software, changed
/etc/dahdi/system.conf to use "hardhdlc" instead of "dchan" - REVERTED
This last change made things worse, so we reverted to the configuration above.
The first four changes appeared to be making things steadily better, but we are
still getting distorted calls and the HDLC errors when the trunk is at or near
capacity, so that could have been wishful thinking on our part.
HI,
Please comment echocanceller in system.conf.
If possible,could you send your SSH to my email:[email protected]?
I'll check it for you.
And you can add my MSN/G-talk:[email protected] to contact with me.
We can now replicate the issue using identical hardware and software config in our development environment, simulating call volume by looping test calls between two of the interfaces. This produces approximately 20 HDLC error messages in a 330 call test.
Have commented out the echocanceller line for the interface, and the test still produces these error messages.
However we have another live system which does not produces these errors. The key difference we can see so far is that the servers producing the error are IBM x3650 M2 79473AG, while the one that works is an older IBM x3650 7979B1A. I am now trying to see if these servers handle interrupts or PCI hardware differently.