How should I set up my distributed environment when applications need to go to shared queue on z/OS?
ColinPaice 120000E7BQ Comments (3) Visits (2476)
How should I set up my distributed environment to provide an HA solution when applications need to go to shared queue on z/OS?
There are three options
Use clients directly into z/OS.
If a client connection to a z/OS queue manager ends, it can reconnect and perhaps go to a different z/OS queue manager. This gives availability to messages on the shared queues. At higher volumes it has a higher CPU cost than using an intermediate queue manager.
Distributed MQ and appliance
With distributed queue manager and the appliance, if the connection stops, the application can connect to a different queue manager, and new traffic flows with only a very small interruption.
Applications must be designed for HA
You need to consider what level of High Availability you need.
Easy test: How many queue managers do your need on distributed if you want to continue to process messages while one queue manager is failing over.
You need at least two. A better solution is to have three or more. If you have two queue managers, and one has a planned shut down, if the other fails over - you have a period when there are no available queue managers. With three queue managers you do not have a single point of failure.
Do I need to use clustering?
With the Shared Queue solution on z/OS you do not need to use clustering. If you are using a non shared queue environment as your back end then clustering gives higher availability.
The hard bit
The implementation of your distributed MQ solution needs careful planning. Some solutions can have a single IP address and port to access multiple queue managers, others solutions need a unique IP address and port for each queue manager.