[VoIP] Firewall logs & status check results

[John R. Covert]

>I suppose the lack of separate timeouts could be corrected

Ahah.  I checked iax.conf.sample in a newer release, and found
a parameter that had not previously existed.

; If we don't get ACK to our NEW within 2000ms, and autokill is set to yes,
; then we cancel the whole thing (that's enough time for one retransmission
; only).  This is used to keep things from stalling for a long time for a host
; that is not available, but would be ill advised for bad connections.  In
; addition to 'yes' or 'no' you can also specify a number of milliseconds.
; See 'qualify' for individual peers to turn on for just a specific peer.
;
autokill=4000

I agreed with the comment that "yes" (2000 ms) is ill-advised,
and set it to 4000.  This would be more like the timeout on an MF
circuit if you hit a stuck register that won't wink back at you.
With this set, the following happened when calling a system that was
not responding:

Jan  1 01:31:39 NOTICE[16627]: chan_iax2.c:2889
  auto_congest: Auto-congesting call due to slow response

But that returned "CONGESTION" rather than "CHANUNAVAIL", and so I
play congestion tone (120 ipm) rather than my "The number you are
calling is currently unavailable."  After removing the autokill (and a
restart of asterisk, not just an IAX2 reload) I get CHANUNAVAIL (and
a proper error message) and the following happens:

Jan  1 01:49:35 WARNING[19159]: chan_iax2.c:1775 attempt_transmit:
  Max retries exceeded to host 206.51.196.18 on IAX2/206.51.196.18:4569-7
  (type = 6, subclass = 1, ts=3, seqno=0)

That takes 30 seconds to happen.  (My comment about it using the
Dial timeout parameter is wrong; it uses timeouts built into the
source.  My dial timeout is 120 seconds.  The source retries four
times in that 30 seconds.)

I should point out that in this case wawinastep.northerntelephone.net
_is_ hung (not totally unreachable), because I CAN ping it:
PING wawinastep.northerntelephone.net (206.51.196.18): 56 data bytes
64 bytes from 206.51.196.18: icmp_seq=0 ttl=44 time=63.953 ms

But it's not responding to IAX2 connections to 14882600.

Since my portal chirps when it begins processing the call, you know
that it's not waiting for more digits.  I've pulled the autokill
back out; I'd rather wait and get the CHANUNAVAIL message.

BTW, the other end can also force a CHANUNAVAIL by issuing a Hangup()
before an Answer().  This is distinct from issuing a Congestion().

Also, I thought Chad Perkins had an automated test and status thing
running, that would only call test numbers explicitly listed as
Milliwatt (or something) in the Member listings.  If we want automated
testing, we should have it driven off of a standardized entry in the
Member Listings, so that (a) you only get tested if you want, and (b)
you can have the test be sent to any number or numbers, since we may
need to test other than just full NXX codes.

/john