HA SIP Server on Windows NLB

[victor]

People,

I need some major help here - I just found out that if you implement high availability with SIP TServers using Windows NLB, it takes 15 to 30 seconds before the backup SIP TServer can be used once primary has crashed.

Genesys told us to use wlb enable/wlb disable command on Windows NLB to activate/deactivate nodes during switchovers. Of course, once you use it, it takes 15+ seconds to switchover... Why would anyone buy HA then?

And of course, I had to find it less than a week before a service-in.

I hope someone has done something with Windows NLB and HA SIP TServers and has already dealt with it...Anyone?

Best regards,
Vic

[Sylvainsjc]

Hi Victor,

HA SipServer with NLB has never worked fine for my customer. We are now looking to deploy with the method in the genesys Sip HA white paper using virtual address IP.

[victor]

Hi, Sylvainsjc,

which part did not work? We are pretty much at service-in now and moving to something else would pretty much ensure project's cancellation...

Best regards (and crying)
Vic

[bublepaw]

Hi Victor,

It is also possible to use Windows Cluster to move IP between two hosts ( although it's to much technology for such task ). Time to move ip from one host to another is about 2 to 3 sec. I didn't test it with SIP server but it's worth of try

Paul

[peters]

Hi Victor,

how did you finally solve this problem? We are also meeting similar problems on NLB, and under hot backup i don't imagine outage of 30 or more seconds. Isn't HA of SIP installed on linux better? Does anybody have some experience?

Thanks, Peter

[victor]

Hi, Peter,

we were able to do it under 5 seconds, with existing sessions not being dropped; however, you will need SIP T/S 7.5 or so. Also, I forgot the reason why, but in the current incarnation, we could not control existing (but not new ones that arrived after switchover) sessions after switchover, so we had to ask client to hang up first. I hope that it was our limitation and not Genesys, even though based one verything I have seen, it was a problem with backup T-Server getting (or failing to get) the right call information from failed T-Server.

Vic

[peters]

Hi Victor,

so this scenario you describe is working using NLB? Just yesterday i realized, why we are getting a strange behavior after HW restart of one node - after joining cluster all SIP traffic is routed to the newly joined node (but running in backup mode), so every new session is dropped, because it doesn't exist on the primary node. It seems the TEAM drivers or switch (port) configuration can be the issue. If we make manual switchover, all existing sessions are kept, but agents logged in genesys desktop get message that "DN is out of service" and have to logout/login again, or wait for about a minute or so.

Peter

[Timur Karimov]

[quote author=victor link=topic=2459.msg15366#msg15366 date=1227842157]
Hi, Peter,

we were able to do it under 5 seconds, with existing sessions not being dropped; however, you will need SIP T/S 7.5 or so. Also, I forgot the reason why, but in the current incarnation, we could not control existing (but not new ones that arrived after switchover) sessions after switchover, so we had to ask client to hang up first. I hope that it was our limitation and not Genesys, even though based one verything I have seen, it was a problem with backup T-Server getting (or failing to get) the right call information from failed T-Server.

Vic
[/quote]
Hi, Vic!

Look at latest Release Note for SIP Server 7.6.x. "Enhanced high-availability support. When operating in a high-availability environment, SIP Server can now synchronize calls that are in ringing state." may be u can try it? but be carefull. when i try in the lab - i brouken my dtmf recongition and can't undestood why =(