callmanager & pg failover

sindhuravi · ‎11-18-2008

We have a testbed setup with follwing components...

icm - 7.2 (duplexed)

cucm - 6.1 ( cluster of 3, 1 pub & 2 sub )

crs - 5.0 (load balancing)

During the failover testing if we stop sub1 and PG1B the contact centre is down. In the configuration the PG1A pointed to sub1 and PG1B pointed to sub2. Please tell me any additional configuration is there or this is limitation of cisco architecture.

Thanks

Sindhuravi

khambetea · ‎11-19-2008

The set up looks correct to me. Two sides of PGs are pointed to the two subs.

I suspect the pguser / password/ hostname / IP settings on PG1A are incorrect.

I would run ICM setup for the PG and confirm. Alternatively you can check jtapi pim registry on PG1A and try failover again. The registry keys are under

HKEY_LOCAL_MACHINE\SOFTWARE\Cisco Systems, Inc.\ICM\\PGA\PG\CurrentVersion\JGWS\jgw1\JGWData\Config

You want to look for jtapiloginstring, jtapipassword and jtapiservice.

If this does not fix issue, it could be an issue with callmanager database replication.

sindhuravi · ‎11-19-2008

Hi thanks for the update...

If i fail the sub1 the PG1B is active and if i fail sub2 the PG1A is active and if i create any user in the cluster it immediately reflecting in all the CUCM. so it seems no issue with pguser & CUCM database.

Thanks

Sindhuravi

cvenour · ‎11-19-2008

>In the configuration the PG1A pointed to sub1 and PG1B pointed to sub2.

This is correctly setup.

>During the failover testing if we stop sub1 and PG1B the contact centre is down.

So you are stopping two of the components, and expecting this to work correctly?

Fault tolerance is designed to avoid single points of failure, through redundancy and fail over.

As the name implies, it provide tolerance for a fault - not necessarily tolerance for multiple simultaneous faults. ICM provides extremely good fault tolerance, but it will not survive what amounts to a catastrophic failure of multiple components.

Based on the information provided, your setup has two communications paths between the CM cluster and ICM:

1. Sub 1 -> PG1A

2. Sub 2 -> PG1b

The solution will be able to continue operation if you lose EITHER a single subscriber, or a single PG side - NOT BOTH (unless you lose both components from the same communication path).

Simply put, the system is behaving as expected.

C.

sindhuravi · ‎11-19-2008

Hi Thanks for your update,

We made a configuration change in the PIM configuration, we added the another CUCM IP seperated by comma(,) during the failover its working as expected, but is this affect any other component and system behaviour.

Since this is not document by cisco we had reverted back to the existing setup.

Please suggest can we proceed on above configuration.

Thanks

Sindhuravi

cvenour · ‎11-19-2008

Unfortunately not. This used to be supported in earlier releases of IPCC, but I understand it was removed because the multiple JTAPI connections resulted in to synchronisation issues between the PG sides and the CTI Manager processes on Call Manager.

As such, Cisco now only support one PIM connection for each CTI Manager, and one CTI Manager connection for each PIM.

C.

d.haeni · ‎05-06-2009

C is completely right - just went through the SRND for 7.5:

http://www.cisco.com/en/US/docs/voice_ip_comm/cust_contact/contact_center/ipcc_enterprise/srnd/75/c7havail.html#wp1060653

d.haeni · ‎05-06-2009

C is completely right. Just went through the SRND for 7.5 and thought I'd share my findings:

http://www.cisco.com/en/US/docs/voice_ip_comm/cust_contact/contact_center/ipcc_enterprise/srnd/75/c7havail.html#wp1060860