| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| WMQ clustering question with multiple WASservers |
« View previous topic :: View next topic » |
| Author |
Message
|
| webspherical |
 Posted: Tue Sep 27, 2005 6:59 am Post subject: WMQ clustering question with multiple WASservers |
 |
|
Acolyte
Joined: 15 Aug 2005
Posts: 50
|
when creating an MQ cluster, isnt it recommended to have the gateway/reposistory QMgr on its own partition? if so, why exactly. thanks
Last edited by webspherical on Tue Sep 27, 2005 2:40 pm; edited 1 time in total |
|
| Back to top |
|
 |
| jefflowrey |
| Posted: Tue Sep 27, 2005 7:00 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
What do you mean, partition?
An LPAR? A disk partition?
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| webspherical |
| Posted: Tue Sep 27, 2005 8:44 am Post subject: |
|
|
Acolyte
Joined: 15 Aug 2005
Posts: 50
|
|
| Back to top |
|
|
| jefflowrey |
| Posted: Tue Sep 27, 2005 9:14 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
There's no particular benefit to putting a gateway qmgr on it's own machine or LPAR. In fact, it increases license costs and etc.
One might do it for different reasons, though. One would be to isolate production from impact. One might be because the gateway qmgr was being used by external customers, and you wanted it in a different network zone than production.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| webspherical |
| Posted: Tue Sep 27, 2005 2:19 pm Post subject: |
|
|
Acolyte
Joined: 15 Aug 2005
Posts: 50
|
question1:So, on the LPAR that has two WAS appservers (using WMQ as the jms provider)
I would have 4 qmgrs, 2 regular QM that service the WAS appservers and the Rep Qmgr and Secondary Rep Qmgr all clustered together.
I would like to take advantage of workload management between the two WAS.
question2: if you were given two LPARs, each with two WAS jvms, to implement workload management why would you create a REP and SREP on each LPAR? wouldn't you want the REP/SREP on a 3rd Lpar to service the other two?
thanks |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Tue Sep 27, 2005 4:46 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
I think of qmgrs being used by WAS instances as bound to that instance.
There's little reason in your scenario to have separate queue managers for your full repositories.
If this is *really* the extent of your MQ network, this is what I would do.
I would create two QMs on each LPAR. One is a Rep and one is a Secondary.
The Rep QM on each LPAR is used by WAS for PUTTING messages. It hosts no qlocals. The Secondary Rep hosts the qlocals, and all MDBs and etc. will GET from these queues.
This will ensure that every message gets workload balanced even in v5.3.
Involving a 3rd LPAR means that you have to pay for a third MQ license for at least one processor. And traffic on a gateway QM is not going to be enough to justify it, and repository traffic (particularly on a two node cluster) isn't going to be enough to justify it.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| webspherical |
| Posted: Thu Oct 20, 2005 4:47 pm Post subject: |
|
|
Acolyte
Joined: 15 Aug 2005
Posts: 50
|
I am somewhat confused at what you are saying. this example shows exactly what I had originally thought. that there is a gateway Qmgr on a separate box..
http://www.redbooks.ibm.com/abstracts/tips0224.html?Open
anyone out there using MQ queue managers behind a gateway with several WAS servers?
thanks |
|
| Back to top |
|
|
| hopsala |
| Posted: Fri Oct 21, 2005 2:30 am Post subject: |
|
|
Guardian
Joined: 24 Sep 2004
Posts: 960
|
webspherical - it seems you had edited your first post, and by the looks of it, though I can't be sure, you had removed data essential to understanding the conversation. I will respond however:
| webspherical wrote: |
| I am somewhat confused at what you are saying. this example shows exactly what I had originally thought. that there is a gateway Qmgr on a separate box.. |
that's not a gateway QM in the redbook, that's just an application QM - serving external applications who send to the clustered queue Q1. If you had another application on your network also sending to Q1, it would have another QM sending directly to QM1 and QM2, as a part of the same cluster. I agree that the name "Messaging Host" is misleading...
In short, I have and do work with WAS clusters and did not use a gateway QM except once, in which there was a Firewall-Security issue involved.
| webspherical wrote: |
| wouldn't you want the REP/SREP on a 3rd Lpar to service the other two? |
I think here lies a misconception - You seem anxious to put full repositories (FR) on a "free" QM, serving no applications; Well, FRs are simply QMs who happen to manage the cluster, they may serve applications as well with no problem at all - managing a cluster is very easy work. There is no need to seperate FRs from the rest of the network.
| jefflowrey wrote: |
The Rep QM on each LPAR is used by WAS for PUTTING messages. It hosts no qlocals. The Secondary Rep hosts the qlocals, and all MDBs and etc. will GET from these queues.
This will ensure that every message gets workload balanced even in v5.3. |
Now this I don't get, maybe because of the edit in the first post - why not just have 1 QM on each machine, one FR one SFR, and be done with it? |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Fri Oct 21, 2005 3:33 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
| hopsala wrote: |
| jefflowrey wrote: |
The Rep QM on each LPAR is used by WAS for PUTTING messages. It hosts no qlocals. The Secondary Rep hosts the qlocals, and all MDBs and etc. will GET from these queues.
This will ensure that every message gets workload balanced even in v5.3. |
Now this I don't get, maybe because of the edit in the first post - why not just have 1 QM on each machine, one FR one SFR, and be done with it? |
I was explaining an optional pattern, that has some value in certain cases, for ensuring that every message that is put gets workload balanced across a cluster properly. It's unneccessary with v6, because you can change the workload balancing, but in v5 as you know the default workload algorithm will always prefer a qlocal over a qcluster. So the workaround is to have a qm that you run your PUTS against, that has no qlocals.
It's not a great workaround, but it's a lot better than writing your own cluster workload exit. And it's a reasonable excuse to have more than one qm on a machine.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| hopsala |
| Posted: Fri Oct 21, 2005 5:41 am Post subject: |
|
|
Guardian
Joined: 24 Sep 2004
Posts: 960
|
| jefflowrey wrote: |
| I was explaining an optional pattern, that has some value in certain cases, for ensuring that every message that is put gets workload balanced across a cluster properly. It's unneccessary with v6, because you can change the workload balancing, but in v5 as you know the default workload algorithm will always prefer a qlocal over a qcluster. So the workaround is to have a qm that you run your PUTS against, that has no qlocals. |
Ah, so you were referring to the case in which WAS appl1 sends to WAS appl2 - and appl2 has two instances and two cluster qlocals? If so, then you're right, it's a very good solution, much better than writing your own exit. I don't think webspherical has a need for this, but i'm glad to have learned of this solution anyway.
| jefflowrey wrote: |
| And it's a reasonable excuse to have more than one qm on a machine. |
Indeed, you got me on that one  |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Fri Oct 21, 2005 5:52 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
| hopsala wrote: |
| Ah, so you were referring to the case in which WAS appl1 sends to WAS appl2 - and appl2 has two instances and two cluster qlocals? |
No, I was referring to essentially bidirectional communication between WAS appl1 and WAS Appl2, with each machine hosting two qms - one for GETTING and one for PUTTING. Each app then has two connections.
This scenario is actually described in some of the scalability planning documentation for WebSphere Commerce.
| hopsala wrote: |
| jefflowrey wrote: |
| And it's a reasonable excuse to have more than one qm on a machine. |
Indeed, you got me on that one |
Well, for v5. For v6, the workload algorithm setting should be used instead, as it vastly simplifies the picture.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| hopsala |
| Posted: Fri Oct 21, 2005 5:56 am Post subject: |
|
|
Guardian
Joined: 24 Sep 2004
Posts: 960
|
| jefflowrey wrote: |
| No, I was referring to essentially bidirectional communication between WAS appl1 and WAS Appl2, with each machine hosting two qms - one for GETTING and one for PUTTING. Each app then has two connections. |
Yes, that's what I meant, Nm! Let's stop before someone hurts himself 
| jefflowrey wrote: |
| Well, for v5. For v6, the workload algorithm setting should be used instead, as it vastly simplifies the picture. |
Truth. Wasn't aware you could change priority to local queues, but it's good to know; haven't been able to find the time to try v6 out yet, I feel somewhat medeivalish... |
|
| 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
|
|
|
|
