Table Of Contents
Switchover Data Mismatch Deferror Prevention Feature Module for Release 5.0 MR 1.1
Revised Date: April 15, 2008
This feature module describes enhancements for Release 5.0 MR1.1 to prevent Oracle data mismatch errors upon switchover. Previously, at EMS switchover, the standby EMS platform was transitioned to active without waiting for the Oracle replication queue to completely drain from the active EMS. To alleviate this, the Session Manager (SMG) now waits 120 seconds when a switchover is initiated to allow the replication queue to drain. Two new commands, Query Oracle Replication Queue on the Active Element Management System (EMS) and Control Element Manager to Switchover have also been added.
Caution If any command response states that the Oracle Replication Queue is broken, contact Cisco TAC.
Query Oracle Replication Queue on the Active EMS
Two new commands permit querying either the status of a job or the number of unpushed calls left in the replication queue.
Job Status Query
This command queries the job status of the Oracle Replication Queue. A user must be logged in to Oracle to execute this command. This example shows that the replication PUSH job is broken. The Oracle Replication Queue is not draining. If the Oracle Replication Queue is broken, it will never drain.
Examplespriems# su - oracle
$ dbadm -r push_job_status
JOB=2 BROKEN=Y FAILURES=0
Query Number of Unpushed Calls
This command queries Oracle for the number of unpushed calls left in the Replication Queue and the estimated time for the Queue to drain. The following example shows that there are 18262 calls in the Replication Queue that may take up to 514 seconds to drain.
Examples$ dbadm -r unpushed_defcall_cnt
Sample response:UNPUSHED_CALLS=18262 ESTIMATED_PUSH_TIME_SECOND=514
Control Element Manager to Switchover
When the CLI Control Element Manager to Switchover command is executed, the system now checks the replication queue for unpushed calls. If the number of unpushed calls is greater than a preset threshold, the control command fails. If switchover is still desired, regardless of the unpushed calls in the queue, the option FORCED=Y is permitted. However, the system now waits 120 seconds before switching over, which allows as many unpushed calls as possible to drain.
Control Element Manager without Forced=Y
This command checks the status of the Oracle Replication Queue before executing a switchover. The following example shows that the control command fails because the Oracle PUSH job is broken. A Database Alarm is given. Contact Cisco TAC for support.
ExamplesCLI>control element-manager id=EM01; target-state=standby-active;
Sample response:Reply: Failure: at 2007-07-19 15:50:42 by optiuserThe replication queue contains 128262 transactions, and it is expectedto take 513 seconds for the queue to drain.
Control Element Manager with Forced=Y
The following example forces the system to switchover. It does not check the status of the Oracle Replication Queue.
ExamplesCLI>control element-manager id=EM01; target-state=standby-active; force=y
Copyright © 2007 Cisco Systems, Inc. All rights reserved.