| Author |
Message
|
| loudcfla |
 Posted: Mon Jun 04, 2007 7:10 am Post subject: MQ Cluster on Z/OS |
 |
|
Acolyte
Joined: 05 May 2002
Posts: 58
|
Right now I have a qmgr on SUN connected to a qmgr on Z/OS and it is connected to a CICS region using CICS MQ adapter. So we do a MQPUT to a queue on SUN via a client, the message comes to Z/OS and runs a CICS transaction and send the reponse back.
If I were to cluster MQ on Z/OS, thus replacing the single qmgr would I still be able to connect to CICS and runs a transaction and send the reponse back?
Page 9 in Chap 1 of V5 QUEUE MANAGER CLUSTERS states "On OS/390 you cannot use clustering if you a using CICS for distributed queueing". |
|
| Back to top |
|
 |
| Vitor |
| Posted: Mon Jun 04, 2007 7:22 am Post subject: Re: MQ Cluster on Z/OS |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| loudcfla wrote: |
| Page 9 in Chap 1 of V5 QUEUE MANAGER CLUSTERS states "On OS/390 you cannot use clustering if you a using CICS for distributed queueing". |
Not in my copy it doesn't, though it does say:
| Quote: |
v On z/OS, if you are using CICS®, you must use the WebSphere MQ mover (not
the CICS mover) to take part in clustering. |
This, I suspect, comes to the same thing. I notice this reference is missing in the v6 version of the manual.
AFAIK there's a 1-2-1 link between queue manager & CICS region. So even under v6 I'd assume you'd need a CICS farm sitting on top of a cluster.
Back in the dim, distant past I remember having a cluster set-up which included a z/OS queue manager & I'm fairly sure that was running CICS but couldn't swear to it. Certainly can't comment on the coding of the transactions running.....
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| cicsprog |
| Posted: Wed Jun 06, 2007 3:51 pm Post subject: |
|
|
Partisan
Joined: 27 Jan 2002
Posts: 347
|
The CICS Mover moves the messages, pre v6, via CICS region connected to TCP Started Task. The CHIN is a separate Started Task that moves the messages via TCP Started Task. The latter the preferred method and IBM chosen one post v5. The CICS Mover is NOT supported in v6. So why even go there unless you already have it configured that way. You will have a conversion step moving to v6 if you do.
I'm confused about your clustering question with z/OS. Given your current configuration – MQ Client connected/sending message to AIX MQM and message sent to z/OS MQM processed by CICS and reply sent back to MQ Client – I don’t see a need for a CLUSTER configuration.
You may need to restate your question more clearly. |
|
| Back to top |
|
|
| loudcfla |
| Posted: Thu Jun 07, 2007 4:56 am Post subject: |
|
|
Acolyte
Joined: 05 May 2002
Posts: 58
|
| Well the problem is since I only have one qmgr processing messages to and from my CICS complex, and only one CICS region dedicated / connected to that qmgr, management feels that they are both single points of failure. |
|
| Back to top |
|
|
| zpat |
| Posted: Thu Jun 07, 2007 5:15 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
Easy enough to have several CICS regions using the same QM and indeed the same queues.
No-one (in their right mind) uses the CICS mover these days.
A single z/OS QM is a SPOF? - yes in theory but I have never seen one fail - I can't see what would make it go down unless the LPAR had hardware problems and that would affect all QMs on the same LPAR.
Personally I wouldn't worry about using only one z/OS QM in any given application, although you might want to isolate some applications in different QMs, rather than attempting clustering which IMHO is more trouble than it's worth.
You could argue that the TCP/IP address space is a SPOF - where do you draw the line? In my view you draw it at the CICS regions - have several of those, that's enough unless you want to have Sysplex and CF. |
|
| Back to top |
|
|
| loudcfla |
| Posted: Thu Jun 07, 2007 5:20 am Post subject: |
|
|
Acolyte
Joined: 05 May 2002
Posts: 58
|
I agree, in all the time I've been here, MQ has never crashed.
How would I connect another CICS to the same MQ, would I just setup another CICS.INITQ for that region? |
|
| Back to top |
|
|
| zpat |
| Posted: Thu Jun 07, 2007 5:40 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
The trouble with standard MQ triggering is that has to refer to a specific CICS region. What happens if that region is not running? I don't know if standard CKTI triggering can share an initiation queue over multiple regions. All our regions have their own initiation queue.
We use our own long-running "trigger monitors" which have outstanding MQGETs (with WAIT) on the application queues. We can run one or more of these "listener or trigger monitors" in one or more CICS regions against the same set of application queues.
It's easy to code something similar yourself, assuming you want to invoke "MQ unaware" CICS transactions upon message arrival.
If your CICS transactions are MQ aware they can simply keep their queues open - no need for triggering.
In production we typically use pairs of CICS regions. This also helps load-balancing and multi-CPU exploitation. |
|
| Back to top |
|
|
|
|
