| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| What happens when a MQ cluster in a HA env. fails over? |
« View previous topic :: View next topic » |
| Author |
Message
|
| rsp |
 Posted: Mon Nov 14, 2011 12:22 am Post subject: What happens when a MQ cluster in a HA env. fails over? |
 |
|
Newbie
Joined: 02 Aug 2004
Posts: 5
Location: Netherlands
|
I am designing an MQ cluster solution to cope with challenging demands. MQ will be used as transport and persisting layer and all connections to MQ will be in client bindings mode. The MQ servers will only just hold the queues, no applications. So each msgs will travel from application to MQ server and then to destination.
After reading the Infocenter V7 documentation and some presentations of the WebSphere Conference, I want to setup 2 MQ clusters: 1 for Entrance, Local and Exit queues, and the other for so called External queues, to divide the load. The first cluster will have to process 24.500 msgs/s (peak) and the other 28.000 msgs/s (peak). The msgs of the first cluster will be 5kB, the other 25kB.
I am planning to use active-passive HA either by a hardware HA solution based on RedHat or using the MultiInstance Qmgrs of WMQV7.
The first MQ cluster will consist of 2 FR's and 7 PR's and the 2nd MQ cluster 2 FR's and 8 PR's to divide the load.
My questions are: do I need to create two exact copies of each MQ cluster, so that per hardware cluster I will have 1 MQ cluster active and the other (exact copy) dormant? In total I will have then 2 MQ clusters active and 2 dormant. Did I forget something in my setup? |
|
| Back to top |
|
 |
| zpat |
| Posted: Mon Nov 14, 2011 12:35 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
With HA, you do not create QMs more than once. On the passive side you only have the necessary entries in mqs.ini and symbolic links in /var/mqm so that when the QM disks fail over the correct references are in place.
With the "old" HA support pac - scripts were provided to do this. With v7 the MI support can do this (whether or not you actually use MI). The actual queue manager objects only exist in one place and messages are recovered over the restart - providing they are persistent. Look into client auto-reconnect as it may be useful to you.
With your volumes, it sounds like you are using non-persistent messages - is this true - it's a key design choice? Your client applications should look at using the "read-ahead" feature to reduce the client channel overhead. Also look at tuning MQ once you have the setup working.
http://www.ibm.com/developerworks/websphere/library/techarticles/0712_dunn/0712_dunn.html |
|
| Back to top |
|
|
| mqjeff |
| Posted: Mon Nov 14, 2011 5:19 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
An mq cluster is completely unrelated to any notions of a highly available queue manager.
A single queue manager is either a member of one or more clusters or not.
A single queue manager is either highly available or not.
These are independent qualities, and have no connection or interrelationship. |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
