[Asterisk-Users] PRI Errors (HDLC Abort (6) on Primary D-channel)

[Andrew McRory]

UPDATE. The circuit has run clean since the 7th. It seems the telco found
a problem after all but they can't tell me what they did...  cause they
think it was fixed as part of another case. Grrr.

Well, I never doubted the T400P too much but I went ahead and bought a
different card just in case. Haven't had to use it so far.

One of the biggest reasons I did not think the Wildcard was at fault is
that the only the LEC PRI span would crash. The other spans stayed up
throughout the ordeal. Also the errors followed the PRI no matter what
port it was plugged into.

For anyone else seeing these errors, here is a checklist of things to look
for:

	1) Run the latest STABLE or CVS release of asterisk/zaptel/libpri
	    If your problems started after an upgrade, revert to the 
	    revision that used to work.
	2) Check and recheck your dialplan for errors. Make backups 
	    often so you can revert to a known good version.
	3) Verify the line framing in /etc/zaptel.conf 
	4) Verify the switch type in /etc/asterisk/zapata.conf
	5) Wildcard must be on its own IRQ. Example:

		[root at phone asterisk]# cat /proc/interrupts
		           CPU0       CPU1
		  0:   52175299          0  local-APIC-edge  timer
		  1:       7513       7309    IO-APIC-edge  keyboard
		  8:          0          1    IO-APIC-edge  rtc
		  9:          0          0   IO-APIC-level  acpi
		 10:          0          0   IO-APIC-level  usb-ohci
		 15:         19          3    IO-APIC-edge  ide1
	---->	 18:  260753631  260932719   IO-APIC-level  tor2
		 20:     815258     827358   IO-APIC-level  ide2, ide3
		 24:          4          8   IO-APIC-level  mvSata
		 27:     836366     849040   IO-APIC-level  eth0
		 28:         38         39   IO-APIC-level  aic7xxx
		 31:    2429141    2497473   IO-APIC-level  eth2
		NMI:          0          0
		LOC:   52176184   52176197
		ERR:          0
		MIS:      12907

	    NOTE: I ran for over 6 months on a system that shared the IRQ 
	    with the Ethernet and USB - just to see what would happen. 
	    It worked very well but when the errors started, this was 
	    the first thing I HAD to fix. Sharing an IRQ with Wildcard is 
	    NOT recommended. Put the card on it's own IRQ! With todays 
	    highly integrated motherboard's, it's getting to where you 
	    can't buy one without an APIC - unless it is a cheapie.

	    NOTE 2: To enable APIC you have to compile it into the kernel 
	    or if running FC1 kernel you must use the SMP version. I am 
	    not sure why APIC is not in the standard FC1 kernel. Planning 
	    to research this further but have not had time to so far.
	6) Motherboard should support APIC (advanced programmable interrupt controller)
	    Use quality motherboard (no el-cheapo chipsets)
	    Use quality memory (run memtest86 to verify)
	    Use quality peripherals
	7) Compile a Kernel from plain sources  
	    ftp://ftp.kernel.org/pub/linux/kernel

	    NOTE: The _Fedora_Core_1_ nptl.2199 kernel available on my 
	    site has been used here with great success. Can't speak for 
	    other precompiled sources since I haven't used them. Also your 
	    hardware may not work as well as mine with this kernel.
	8) Make sure you have good power / power supply. Run a UPS!
	9) If a multiport card, move the PRI/T1 span to another port to 
	    see if the errors follow.
	10) If you have extra equipment to install on an unused span, 
	    configure it and see if it crashes at the same time the
	    problem circuit does. If all circuits crash at the same time 
	    that would indicate a hardware problem such as IRQ or noise on 
	    the PCI bus (bad motherboard or PCI controller??).

If all the above checks out these Red Alarms and HDLC errors are most
likely due to problems in provisioning the circuit. At this point, become
the Telco's biggest squeaky wheel or wait weeks for them to fix it by
accident, if at all.

Once more here is my configuration and it works :-)

			(T400P) <--> <ASTERISK> <--> SIP/VoIP/etc.
			|
	LEC-PRI <----->	(Port1)
			(Port2) <---> <Max 40000 / V.90 Dial-up users>
			(Port3) <---> <Microcom isPorte / Fax Test>
			(Port4) <---> <unused / test>

Thanks to all who offerd help!! I hope to not have to post about this 
error again!

-- 
Andrew McRory - President/CTO 
Linux Systems Engineers, Inc. - http://www.linuxsys.com
Located in beautiful Tallahassee, Florida
Office  850-224-5737
Office  850-575-7213
Mobile  850-294-7567