How to configure IBM MQ as a HA cluster? - ibm-mq

Question

I want to create a high availability cluster. But i cant find any steps to create one from the IBM documentation.
I followed this guide to create the cluster.
Both QM are deployed with docker-compose:
version: '3.7'
services:
london:
build:
context: ./config/london
environment:
LICENSE: "accept"
MQ_QMGR_NAME: "QM1"
MQ_ENABLE_METRICS: "true"
ports:
- 9443:9443 # web view
- 9157:9157 # metrics
newyork:
build:
context: ./config/newyork
environment:
LICENSE: "accept"
MQ_QMGR_NAME: "QM2"
MQ_ENABLE_METRICS: "true"
ports:
- 9553:9443 # web view
- 9158:9157 # metrics
depends_on:
- london
Dockerfile for both images:
FROM ibmcom/mq
COPY init.mqsc /etc/mqm/20-init.mqsc
Here is the config for the london MQ instance:
* cluster config
ALTER QMGR +
REPOS(INVENTORY) +
PSCLUS(ENABLED)
DEFINE LISTENER(LONDON_LS) +
TRPTYPE(TCP) +
CONTROL(QMGR)
DEFINE CHANNEL(INVENTORY.LONDON) +
CHLTYPE(CLUSRCVR) +
TRPTYPE(TCP) +
CONNAME('london(1414)') +
CLUSTER(INVENTORY) +
DESCR('TCP Cluster-receiver channel for queue manager LONDON')
DEFINE CHANNEL(INVENTORY.NEWYORK) +
CHLTYPE(CLUSSDR) +
TRPTYPE(TCP) +
CONNAME('newyork(1414)') +
CLUSTER(INVENTORY) +
DESCR('TCP Cluster-sender channel from LONDON to repository at NEWYORK')
SET CHLAUTH('INVENTORY.LONDON') +
TYPE(QMGRMAP) +
QMNAME(QM2) +
USERSRC(CHANNEL) +
ADDRESS('*')
and here for the newyork instance:
* cluster config
ALTER QMGR +
REPOS(INVENTORY) +
PSCLUS(ENABLED)
DEFINE LISTENER(NEWYORK_LS) +
TRPTYPE(TCP) +
CONTROL(QMGR)
DEFINE CHANNEL(INVENTORY.NEWYORK) +
CHLTYPE(CLUSRCVR) +
TRPTYPE(TCP) +
CONNAME('newyork(1414)') +
CLUSTER(INVENTORY) +
DESCR('TCP Cluster-receiver channel for queue manager NEWYORK')
DEFINE CHANNEL(INVENTORY.LONDON) +
CHLTYPE(CLUSSDR) +
TRPTYPE(TCP) +
CONNAME('london(1414)') +
CLUSTER(INVENTORY) +
DESCR('TCP Cluster-sender channel from NEWYORK to repository at LONDON')
SET CHLAUTH('INVENTORY.NEWYORK') +
TYPE (QMGRMAP) +
QMNAME(QM1) +
USERSRC(CHANNEL) +
ADDRESS('*')
I want to have a topic in which my applications can write/read from regardless of the QM they are connected to.
Is this even possible and if so could you share the MQSC commands or a link to a example setup.

Answer 1

MQ clustering is about load balancing connectivity between queue managers.
You could have for example a clustered queue that is defined on LONDON and NEWYORK, this could be PUT(ENABLED) on both (hot/hot) and message sent to that queue by cluster members will be in general round robin between the two instances.
If you want it to be hot/warm, then you would PUT(DISABLE) one of the two instances for example LONDON, then you would direct your application to connect and read from NEWYORK, if NEWYORK went down, you could PUT(ENABLED) the queue on LONDON and message from that point forward will go to LONDON and you can direct your application to LONDON (note that any message that were sitting on the queue on NEWYORK will only be available when NEWYORK is recovered. MQ does not replicate the messages to both queue managers.
With pub/sub it is slightly different in that all queue managers in the cluster will know about clustered topics, and if any subscriber on any QM in the cluster subscribes to that TOPIC a proxy subscription will be made by the subscriber QM, if any publishes happen to that topic on any queue manager a copy will be directed to each local subscriber as well as any QMs with a proxy subscription in place, the subscriber QM will then deliver the message to any subscribers local to it (so only one proxy sub even if multiple subscribers on that same QM).
If you had non-durable subscribers they could then subscribe to the topic on any QM and receive messages published after the subscription is created but not before.
If you required that you receive messages when the subscriber is not connected you would need a durable subscription and that would be tied to the specific queue manager it was made on.
You may want to look at RDQM which does provide replication between nodes, but is not compatible with containers.