UCCE - Split Brain concept – Agent PG A side –Ideal

Ramamoorthy Shanmugam · ‎05-14-2017

Dear All,

We are running on UCCE 10.5.x with Duplex setup . last we had major outage due to Wan link failure on SIDE–B Data center is not accessible .

During the time Majority of the PG’s are active in B side . During this time IVR is played and while taking agent Transfer calls are disconnected in Queue to skill group node .

Agent PG – A is idea , We have disable the Router configuration changes and disabled on Rogger A & Agent PG A private link and cycle the process .

set the HKEY_LOCAL_MACHINE\Software\Cisco Systems, Inc.\ICM\<instance_name>\RouterA\Router\CurrentVersion\Configuration\Global\DBMaintenance key to 1

After cycle also Agent PG – A is idle .

Please let us know during this scenario how to activate A side ICM servers and deactivate B side process . or how to bring in service for A side .

Ram.S

Regards,
Ram.S

geoff · ‎05-14-2017

Am I correct in assuming that the WAN failure has taken out both the private and public links?

Do you have any access to the B side in order to prevent calls coming in there?

When you are forced to run in "split-brain" mode on the Central Controller, you want to prevent the Loggers from acting independently, each writing data for the calls, because you can then have significant problems when you try to bring the system back in duplex mode if there are clashes between Recovery Keys between the sides.

It sounds like you want to ignore side B and run everything on side A. This certainly should work.

Is the agent PG not coming up because the PIM will not go active and the reason it will not is that your JTAPI connection is misconfigured and pointing to a Subscriber at the B side?

If that is not the case and the side A PIM is connecting to a CTI Manager process at the A side, then you may have something else misconfigured on the PG to do with the Network Connections to the Router.

You would have (and should have) sorted all of this out before going live with a correct and complete Functional and Failover test plan.

In principle, you should be able to do what you want. You need to look at the logs and determine why the PG at the A side will not go active.

Regards,
Geoff

Ramamoorthy Shanmugam · ‎05-14-2017

Hi Geoff,

Am I correct in assuming that the WAN failure has taken out both the private and public links? : Yes both the link went down

Do you have any access to the B side in order to prevent calls coming in there? : No I tried Private and Public IP are not reachable

When you are forced to run in "split-brain" mode on the Central Controller, you want to prevent the Loggers from acting independently, each writing data for the calls, because you can then have significant problems when you try to bring the system back in duplex mode if there are clashes between Recovery Keys between the sides.

It sounds like you want to ignore side B and run everything on side A. This certainly should work. : Yes – I need work in A side

Is the agent PG not coming up because the PIM will not go active and the reason it will not is that your JTAPI connection is misconfigured and pointing to a Subscriber at the B side? Your right we have 3 CUCM node , 1 PUB , 2 SUB : A side we have given SUB 1 IP and B side given SUB 2 IP with the same PG application user name and password

If that is not the case and the side A PIM is connecting to a CTI Manager process at the A side, then you may have something else misconfigured on the PG to do with the Network Connections to the Router.

You would have (and should have) sorted all of this out before going live with a correct and complete Functional and Failover test plan. : This system is working for past three years . Recently we have migrated to 10.5 and every three month we are doing for Failover and resilience test on both the side for Components and Services level . we never faced any issues for service level .

In principle, you should be able to do what you want. You need to look at the logs and determine why the PG at the A side will not go active. : We are giving side A preference for all the Router, PG’s .

Also few times central controller and PG’s services are failed for some reason , that time other side services are taking control and we never faced any business impact .

Ram.S

Regards,
Ram.S

geoff · ‎05-15-2017

I am sure by now you have opened a TAC case to get assistance, but if you feel like uploading a couple of log files from the A side Agent PG, we could take a look. pgag, pim1, jgw1, mds.

Regards,
Geoff