[OpenSIPS-Users] crashing in 2.2.2
Richard Robson
rrobson at greenlightcrm.com
Fri Mar 3 06:28:21 EST 2017
I've revisited the gateway failover mechanism I had developed in order
to re route calls to the next gateway on 500's due to capacity on the
gateways we are using.
we have 3 gateways from one carrier and one from another. The 3 have 4
cps and will return a 503 or 500 if we breach this. The single gateway
from the other carrier has plenty of capacity and should not be a
problem so we want to catch this . and route to the next gateway.
We are counting the CPS and channel limits and are routing to the next
gateway if we exceed the limit set, but There are still occasions where
a 5XX is generated, which results in a rejected call.
We want to stop these rejected calls and therefore want to implement the
failover mechanism for the 5XX responses. For 6 months we have been
failing over if we think the counts are to high on any one gateway
without a problem. But when I implement a failover on a 5XX response
opensips starts crashing.
It's difficult to generate artificial traffic to mimic the real traffic,
but I've not had a problem with the script in testing. Last night I
rolled out the new script but by 09:15 this morning opensips started
crashing 10 times in 5 minutes. This was as the traffic ramped up. I
rolled back the script and it restarted OK and has not crashed since.
Therefore the Failover Mechanism in the script is where the crash is
happening
Core dump: http://pastebin.com/CqnESCm4
I'll add more dumps later
Regards,
Richard
this is the failure route catching the 5XX
failure_route[dr_fo] {
xlog (" [dr] Recieved reply to method $rm From: $fd, $fn, $ft,
$fu, $fU, $si, $sp, To: $ru");
if (t_was_cancelled()) {
xlog("[dr]call cancelled by internal caller");
rtpengine_manage();
do_accounting("db", "cdr|missed");
exit;
}
if ( t_check_status("[54]03")) {
route(relay_failover);
}
if ( t_check_status("500")) {
route(relay_failover);
}
do_accounting("db", "cdr|missed");
rtpengine_manage();
exit;
}
This is the route taken on the failure
route[relay_failover]{
if (use_next_gw()) {
xlog("[relay_failover-route] Selected Gateway is $rd");
$avp(trunkratelimit)=$(avp(attrs){s.select,0,:});
$avp(trunkchannellimit)=$(avp(attrs){s.select,1,:});
####### check channel limit ######
get_profile_size("outbound","$rd","$var(size)");
xlog("[relay_failover-route] Selected Gateway is $rd
var(size) = $var(size)");
xlog("[relay_failover-route] Selected Gateway is $rd
avp(trunkcalllimit) = $avp(trunkchannellimit)");
xlog("[relay_failover-route] Selected Gateway is $rd
result = ( $var(size) > $avp(trunkchannellimit))");
if ( $(var(size){s.int}) >
$(avp(trunkchannellimit){s.int})) {
xlog("[relay_failover-route] Trunk $rd exceeded
$avp(trunkchannellimit) concurrent calls $var(size)");
route(relay_failover);
}
} else {
send_reply("503", "Gateways Exhusted");
exit;
}
##### We need to check Rate Limiting #######
if (!rl_check("$rd", "$(avp(trunkratelimit){s.int})",
"TAILDROP")) { # Check Rate limit $avp needs changing
rl_dec_count("$rd"); # decrement the counter since
we've not "used" one
xlog("[ratelimiter-route] [Max CPS:
$(avp(trunkratelimit){s.int}) Current CPS: $rl_count($rd)] Call to: $rU
from: $fU CPS exceeded, delaying");
$avp(initial_time)=($Ts*1000)+($Tsm/1000);
async(usleep("200000"),relay_failover_delay);
xlog ("Should not get here!!!! after async requst");
} else {
xlog ("[relay_outbound-route] [Max CPS:
$avp(trunkratelimit) Current CPS: $rl_count($rd)] Call to: $rU from: $fU
not ratelimited");
}
t_on_failure("dr_fo");
do_accounting("db", "cdr|missed");
rtpengine_manage();
if (!t_relay()) {
xlog("[relay-route] ERROR: Unable to relay");
send_reply("500","Internal Error");
exit;
}
}
--
Richard Robson
Greenlight Support
01382 843843
support at greenlightcrm.com
More information about the Users
mailing list