FAQ: High Availabilty using Multi-Instance Message Broker
nsrao 120000KVEK Visits (12239)
The introduction of the multi-instance(MI) concept into the Websphere MQ and WebSphere Message Broker products helped several clients to make use of this freely available high availability feature in their infrastructure. The earlier cat 2 HA support packs IC91 and MC91 became irrelevant with this new functionality in the products. Thus the multi-instances feature made the existing HA user's job very simple.
There have been a number of articles and presentations on how to set up multi-instance queue managers and message brokers. I have included some interesting URLs of this pre-exsting information, so I won't repeat the same information in this post:
But, I've received some feedback and questions on the above, so I thought I would summarize some of the questions for which the answers are not very straight forward to grasp from the above URLs.
Q1. How does the Multi-Instance failover help for my webservice fail over?
Ans: MI Broker failover happens via MQ transport and other transports like SOAP or JMS failovers and automatic reconnections are not currently supported.
Q2. How can I achieve failover of my webservice flow?
Ans: For the message flow which use JMS or SOAP nodes would need a network dispatcher in front of Multi-instance message brokers.
Q3. Can I develop Java™ Servlets for failover reconnect?
Ans: HTTP Servlets won’t help in connecting to the host with standby node when it gets active during the failover.
Q4. My Shared-library files are not accessed after broker fail-over; why?
Ans: Shared library folder on /var/mqsi is not failed over to stand-by node during the failover process. The best practice is to have a copy of the library files in both nodes or have this folder created in the shared NFS filesystem, so that is available both active and standby nodes.
Q5. Are there any added advantages with multi instances on top of HACMP or commercially available Cluster Software?
Ans: Generally speaking Multi-Instance(MI) Brokers provide only a subset of the capabilities that commercial cluster software like HACMP.
Q6. Can I set up active-active multi-instaces?
Ans: MI Brokers are aimed for a failover HA, but NOT for active-active and/or load balancing scenarios.
Q7. Can I use any filesystem for shared folder?
Ans: There are three fundamental requirements that a shared file system must meet:
Q8. Is it correct that multi-instance queue managers had to run on domain controllers?
Ans: In WebSphere MQ for Windows version 7.1, you can run multi-instance queue managers on Windows servers or workstations that are part of a domain. In version 7.0.1, multi-instance queue managers had to run on domain controllers only.
Q9. Can I create different userIDs on the two failover nodes of Multi-instances?
Ans: The UID and GID of mqm and wmbuser should be same on both machines before WebSphere MQ installation. If they are different then change them to be the same and reboot the system to reflect the changes.
Q10. How do I remove multi-instance broker and queue managers?
Ans: The following steps in this order would remove the MI broker and QMs successfully:
Q11. What files/folders should be on sharedFS and which ones should be on localFS?
Ans: Directories local to a broker registry: contains HASharedWorkPath, Plugins and Shared classes
Directories shared by MI Qmgrs:
qmgrs - contains qmgr data
logs - contains qmgr logs
Directories shared by MI Brokers:
mqsi - contains components (config, httplistener, locks, repository, stderr, stdout), regi
Q12. How do I backup and restore MI brokers?
Ans: Backup a multi-instance broker using following command:
mqsibackupbroker MI_BK -d <bac
Restore a multi-instance broker using following command:
mqsirestorebroker MI_BK -d <bac