Resource Manager NLB HA behaviour

[bublepaw]

routeset option in sip and vrmrecorder section should contain FQDN of active-active cluster - in such case switchover between two RM works without any further configuration. For active-active RM based only on ip addresses You have to configure transport.staticroutelist option which should contain both RM ip separated by comma.

[Kubig]

So, for active-active RM cluster based on IP address list only is necessary to set transport.staticroutelist to comma separated IP addresses of each RM in cluster and in routeset (sip,vrmrecorder) one of these IP's or can be empty? I have always deployed and configured with both options. And all worked fine (announcement,vxml and rec services).

Timur has right that the function is not described very well in actually doc and some parts is missing.

[Timur Karimov]

[quote author=bublepaw link=topic=7736.msg34379#msg34379 date=1372690518]
routeset option in sip and vrmrecorder section should contain FQDN of active-active cluster
[/quote]
WOW?! Wait a second! That's a "FQDN of active-active cluster" ?! In active-active cluster we didnt' have any shared IP or FQDN, each RM have independent IP and FQDN name!

[bublepaw]

That is why I mentioned that there are 2 ways of deploying active-active cluster:

- each RM has it's own ip address and in all configurations we enter both ip addresses where appropriate (contact-list in SIPS, routset and staticroutelist)
- each RM has it's own ip address but we reference RM pair using FQDN which resolves to both ip addreses. We use it in routeset and contact options

Another difference in using DNS is fact that we can use different port for each instance while with ip addresses we need to use same port for both instances of RM.

[Timur Karimov]

Hmm,
you mean use only IP for that options wich can use the more than once address, but use the FQDN for options which can contain only one address?
Interesting trick. But i see a little problem. In all service wich work with DN we have the OOS detection. And standart behavior for SIP - mark DN as out-of service and don't use it while oos timer is not counted. Did you already test such config ? And for me the most impotant question - did you ever try it with third-party recorer, Zoom for example?

[Kubig]

Yes, that is what I was thinking. I have real experience with using it (a-a cluster based on IP only) with ZOOM recorder and it works fine. I mean better than active-standby mode. In active-standby mode sometimes the "switchover" takes a long time before the backup node became to primary node..

[Timur Karimov]

Fanx, Kubig! You words is make the hope grew in my soul =)

[Timur Karimov]

Looks like it's work =)) Fanx again to all in this topic!

[Timur Karimov]

ROFL =)))) Look at this answer from official  Genesys Support:
"From provided configuration I can see that you are running RMs in active-active mode.
Please note method required external load balancer which owns a virtual IP address that is used to forward requests to the active cluster"

[bublepaw]

Ignore it and point them to manual :>. What problem You had?

[Timur Karimov]

Sip server is not evenly distributes the call on RM.  And the main problem - RM have the "stalled" call which leads to the port leak

[bublepaw]

Try setting option  sip-enable-rfc3263 to true to solve distribution issue. As mentioned previously feature was designed to work with FQDN's and it works much better in such setup

[Kubig]

Additional question : It is necessary or recommended to add RM in active-active cluster to HotStandby mode on application object under CME or GA?

[Timur Karimov]

[quote author=bublepaw link=topic=7736.msg34476#msg34476 date=1373440836]
Try setting option  sip-enable-rfc3263 to true to solve distribution issue. As mentioned previously feature was designed to work with FQDN's and it works much better in such setup
[/quote]
already enabled

[bublepaw]

[quote author=Kubig link=topic=7736.msg34477#msg34477 date=1373443550]
Additional question : It is necessary or recommended to add RM in active-active cluster to HotStandby mode on application object under CME or GA?
[/quote]

No - RM pair should be configured as two separate applications without any HA settings