9 Aug
2022
9 Aug
'22
4:35 p.m.
Hi Experts,
We have been using Kamailio in an Active/Standby Pair(with Keepalived under
the rugs moving the single Virtual IP to access the Active Kamailio) for
sometime now. *Kamailio also acts as a Registrar for our webrtc endpoints. *
It has been serving the purpose pretty well and now we have a requirement
where we need to be syncing the Registration DB between Two Pairs of
Kamailios.
Kamailio-Active(*Pair 1*) -------- Kamailio-Standby(*Pair 1*)
||
Kamailio-Active(*Pair 2*) -------- Kamailio-Standby(*Pair 2*)
We generally keep the counterpart in the same Pair as a notifier(
*modparam("dmq",
"notification_address", "sip:PEER1_IP:5060")* ) so as to sync the
dialogs
and the userloc data too( *modparam("dmq_usrloc", "enable", 1) * ).
In order to achieve the said requirement with the other Pair, we added
another "*notification_address*" in the kamailio cfg. At this point, we ran
into weird issues.
*1. *With Kamailio ver *5.3.2, *the subsequent *notification_address *line
in the cfg file, seemed to be overriding the previous one. Hence we see
only the latter peer in the dmq list nodes.
Example:
modparam("dmq", "notification_address",
"sip:*172.27.45.77*
:5090")
modparam("dmq", "notification_address",
"sip:*172.27.45.200*
:5090")
In this case, the "*kamcmd dmq.list_nodes*" would show the local
Machine and 172.27.45.200 as the only nodes in the output, ie *172.27.45.77
is not showing up at all,* which is problematic, since the local
machine(172.27.45.243 in our case) would not been able to send any dmq sync
info to its peer(172.27.45.77) in the same Pair.
To see if the above issue might have been addressed in later release, we
upgraded to the latest Kamailio ver *5.6.0*
*To our respite, the above issue no longer exists in the new version*(though
not sure which immediate release after v 5.3.2 it would have been initially
fixed.)
This is where we have a new issue explained below:
*2. *The registration data does get synced to the peer Kamailio in the same
Pair, and also to the Kamailio instances in the other Pair. However, the
*Socket* Parameter in "*kamctl ul show*" output shows *[not set] *even on
the side where the websocket connection actually exists.
[root@localhost ~]# *kamctl ul show *
{
"jsonrpc": "2.0",
"result": {
"Domains": [{
"Domain": {
"Domain": "location",
"Size": 1024,
"AoRs": [{
"Info": {
"AoR": "9008077221",
"HashID": 1952082106,
"Contacts": [{
"Contact": {
"Address": "sip:Harneet_qifir@172.24.58.210",
"Expires": 159,
"Q": -1,
"Call-ID": "vfli2uv8du3ppda73q5ppe",
"CSeq": 106,
"User-Agent": "EngageDigital",
"Received":
"sip:172.27.44.252:60070;transport=ws",
"Path": "[not set]",
"State": "CS_NEW",
"Flags": 0,
"CFlags": 0,
* "Socket": "[not set]",
<<<<<<<<<<<<<<<<<<<<<<*
"Methods": 7071,
"Ruid": "uloc-62f26e3b-2677-1",
"Instance":
"<urn:uuid:4fbd3e96-b4de-497b-886d-1ca8ffa016a4>",
"Reg-Id": 1,
"Server-Id": 0,
"Tcpconn-Id": -1,
"Keepalive": 0,
"Last-Keepalive": 1660063892,
"KA-Roundtrip": 0,
"Last-Modified": 1660063892
}
}]
}
}
],
In order to confirm that the socket actually exists on this Kamailio
instance, I am pasting the below outputs from the same machine, where ws
dump and even the native netstat confirms that.
[root@localhost ~]# *netstat -tunelap | grep 60070*
tcp 0 0 172.27.45.199:8080 172.27.44.252:*60070*
*ESTABLISHED* 994 17322355 16230/kamailio
[root@localhost ~]# *kamcmd ws.dump*
{
connections: {
1: ws:172.27.44.252:*60070* -> ws:172.27.45.199:8080 (state: *OPEN*, last
used 24s ago, sub-protocol: sip)
}
info: {
wscounter: 1
truncated: no
}
}
We do need to distinguish the actual Kamailio instance where the websocket
connection actually exists, so as to route the call ahead from the same
instance, or if it does not exist(and it's merely a sync'ed registration
data received over DMQ Channel), then the Kamailio should route the call
ahead to the Kamailio instance in the other Pair, which can then route it
ahead to the Registered webrtc endpoint. We were hoping to use the Socket
Parameter output, but for the said problem, unable to use the same as an
indicator.
*So what would be the best way to identify which Kamailio has the websocket
connection with the Actual endpoint*? Should we rely on the output of
netstat or ws.dump to infer that? I mean this needs to be done in
kamailio.cfg for each call, so want to know the best way, or if there is
completely different approach that can be suggested?
Apologies for the long email, but any pointers will be much helpful.
Thanks & Regards,
Harneet Singh
--
"Once you eliminate the impossible, whatever remains, no matter how
improbable, must be the truth" - Sir Arthur Conan Doyle