MQSeries.net :: View topic

amit_talpade_star · Posted: Thu Dec 08, 2005 3:08 pm Post subject:

We have same cluster queue names on two queue managers SF_QM1, and SF_QM2 for load balancing

The network interface for one of the queue manager(SF_QM1) is intentionally stopped to check for failover.

We have observed that it takes over 5-7 minutes for other cluster queue managers to detect this and stop sending the requests to the failed queue manager

In this process, we are losing some messages. The messages being sent are not persistent, however the expectation would be that messages should be silently and immediately routed to the available queue manager where a queue of the same name exists.

Is there a way that the cluster channels can be configured to avoid such a behavior where messages are lost and routed to other running queue manager

A channel configuration on one of the queue manager has been provided for your reference.

***********queue manager details****************

ALTER QMGR +
DEADQ('SYSTEM.DEAD.LETTER.QUEUE') +
PERFMEV(ENABLED) +
FORCE

DEFINE CHL(SYSTEM.ADMIN.SVRCONN) CHLTYPE(SVRCONN) REPLACE

**********Cluster Channels***********************

DEFINE CHANNEL ('TO.BRK_QM1') CHLTYPE(CLUSSDR) +
CONNAME('xxx.xxx.xxx.xxx(1414)') +
TRPTYPE(TCP) +
CLUSTER('MUA_CLUSTER') +
DISCINT(0) +
DESCR('TCP Cluster-sender channel from SF1 to MB1') +
NPMSPEED(FAST) +
MAXMSGL(104857600) +
REPLACE

DEFINE CHANNEL ('TO.SF_QM1') CHLTYPE(CLUSRCVR) +
CONNAME('xxx.xxx.xxx.xxx(1414)') +
TRPTYPE(TCP) +
CLUSTER('MUA_CLUSTER') +
DISCINT(0) +
DESCR('TCP Cluster-receiver channel for SF1') +
MAXMSGL(104857600) +
NPMSPEED(FAST) +
MRRTY(2) +
MRTMR(10) +
REPLACE