Kamailio 3.0.4: dispatcher module request - sr-users

24 Feb 2011

Hello list,

I have a question to the dispatcher module in Kamailio version 3.0.4.

In my integration I use this module for distributing calls to a set of
gateways (with the round robin algorithm). Sometimes a gateway does not
react or send back a negative response. In that case I use the function
'ds_mark_dst("p")' in the failure_route and mark the current
destination
with the flag 'probing'. I have set the threshhold value for probing mode
to '10'. This means that a gateway is effectively set to probing mode only
when it was marked with the flag 'probing' for 10 times. However,
sometimes a gateway is sending back a negative response for e.g. 5 times
and is afterwards working fine - without any problems. A few hours later
(e.g.) it is sending a negative reply again for one time. This means: the
counter for the probing_flags is increasing every time when the function
ds_mark_dst("p") is executed (in case that the process was not restarted
in the meantime!). So the problem is, that the counter is reaching the
threshhold value at any time and at that moment the gateway is set to
probing mode. Even if the gateway has sent a negative reply for only one
time (at that moment) - but according the history the proging_flag was set
several times.

What I miss in this module is a function to _reset_ that flag counter from
a REPLY_ROUTE. Note: I have to reset the counter, because the gateway does
not yet SIP OPTIONS requests and therefore it is set to 'probing' all time
until it is manually set back to active via MI command or after a process
restart. According the README file (and practical experience) the function
ds_mark_dst() is executable only within the failure_route. However, when I
want to set the flag ("a") for the current destination I will do it within
a reply_route, but not in the failure_route (because I have received a
positive response and not a negative one). From my point of view it does
not make sense setting the 'active' flag for a destination from within a
FAILURE route.

Does anybody have an idea, how the 'flag-counter' could be reset (from
within the script)? I do not want to use a MI command.....

configuration excerpt:

[...]
modparam("dispatcher", "db_url",
"mysql://openser:openserrw@localhost/openser")
modparam("dispatcher", "table_name", "dispatcher")
modparam("dispatcher", "dst_avp", "$avp(AVP_DST)")
modparam("dispatcher", "grp_avp", "$avp(AVP_GRP)")
modparam("dispatcher", "cnt_avp", "$avp(AVP_CNT)")
modparam("dispatcher", "ds_ping_from",
"sip:proxy@192.168.37.87")
modparam("dispatcher", "ds_probing_mode", 0)
modparam("dispatcher", "ds_probing_threshhold", 10)
modparam("dispatcher", "ds_ping_interval", 30)
modparam("dispatcher", "flags", 2)
modparam("dispatcher", "force_dst", 1)
[...]
     if (!ds_select_dst("1", "4")) {
          sl_send_reply("404", "No destination (disp)");
          exit;
     }
     t_on_failure("failureroute");
     t_on_reply("replyroute");
[...]
failure_route[failureroute] {
[...]
     if (t_check_status("[45][08]0") || (t_branch_timeout() &&
!t_branch_replied())) {
          ds_mark_dst("p");
          if (!ds_next_dst()) {
               t_reply("404", "No destination (disp)");
               exit;
          }
          t_on_failure("failureroute");
          t_on_reply("replyroute");
     }

[...]
}
onreply_route [replyroute] {
     if (t_check_status("180|200") {
          # nice to have......
#          ds_mark_dst("a");
     }
}

Thanks in advance!

regards,
Klaus