| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| HW Cluster + Active/Active config in Unix server |
« View previous topic :: View next topic » |
| Author |
Message
|
| TTB |
 Posted: Thu Jul 13, 2006 1:01 am Post subject: HW Cluster + Active/Active config in Unix server |
 |
|
Newbie
Joined: 12 Jul 2006
Posts: 2
Location: HK
|
Dear All,
Please help. As a newbie, I am hoping to gain better knowhow on high-availability clustering in Unix server, especially on the following:
- HW Cluster
- Active/Active config.
Having read the MC91*, article published by Mark Hiscock and Simon Gormley**, and messages in this forum, however there are still unclear MQ config concepts, and I would be grateful if anybody can help sharing some insight/guidance.
In short I have the following hw/sw components identified:
2 Unix Server (called svr1 and svr2)
QM on each server (called svr1_qm and svr2_qm)
VCS for clustering service
A) Assuming two different application would be deployed in this Active/Active config (svr1 host app1 and svr2 host app2 initially)
1) Must there be 2 shared disk (as only 1 active resource group can access a disk at a time, therefore under Active/Active config, 2 active resource group need 2 disk)?
2) Therefore does an "Active/Active config" mean two "Active/Standby config" running in parallel?
B) Assuming the two application instances deployed on the two server are the same application (each svr1 and svr2 host separate instance of the same app):
3) Can QM svr1_qm and svr2_qm and their respective queue have the same name (since the two instances of app running on svr1 and svr2 are the same)? Or must svr1_qm and svr2_qm, and their respective queue have an unique name?
4) How would the external message target this Active/Active environment? Must I have additional server, meaning svr3, svr3_qm and cluster queue to form MQ Clustering? Otherwise how can one publish a single queue for gathering external message (to achieve a single entry point for external system)?
5) Provided the two QM or two queue have the same name, if svr1 fail over to svr2, how can the app distinguish which queque to read msg from?
Thank you in advance.
* MC91: High Availability for WebSphere MQ on UNIX platforms Version 6.0
** Understanding high availability with WebSphere MQ by Mark Hiscock and Simon Gormley |
|
| Back to top |
|
 |
| PeterPotkay |
| Posted: Thu Jul 13, 2006 5:00 pm Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
1. Yes, 2 disks needed. 4 If you want the QM logs on a separate disk than the QM data.
2. I guess that's a way of thinking about it.
3. The QMs must be named differently. But they can host identical queues.
4. If you want to load balnce messages between the QMs, put them into an MQ clusters. At MQ 5.3, you need a third QM that does not host the target queues to round robin the messages. In MQ 6.0, you don't need that 3rd QM, as you can set queue properties to make it such that you load balance even if there is a local queue on the connected QM.
5. The 2 QMs must be named differently.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| TTB |
| Posted: Thu Jul 13, 2006 10:53 pm Post subject: |
|
|
Newbie
Joined: 12 Jul 2006
Posts: 2
Location: HK
|
Dear PeterPotkay,
Thanks for your reponse.
Having ensured the two QM must have different name, could you please (or anybody) kindly tell me the following:
1) From the statement:
"4. If you want to load balnce messages between the QMs, put them into an MQ clusters. ..."
given the goal is to provide a single access point for external application to target this Active/Active environment, is there any option apart from MQ clustering?
2) Using the following setting:
2 Unix Server (called svr1 and svr2)
QM on each server (called svr1_qm and svr2_qm)
VCS for clustering service
Active/Active config, one application on each server (called svr1_app and svr2_app, two instances of the same application hosted on svr1 and svr2 respectively)
2.1) If svr1 fails, svr1_qm fails over to svr2. Must svr1_app fails over to svr2 as well?
2.2) If not, then the following would all be hosted in svr2:
svr1_qm - due to failover from svr1
svr2_qm - originally exist in svr2
svr2_app - originally exist in svr2, svr1_app remains at the failed svr1
Can svr1_qm and svr2_qm be configured to provide a single access point for svr2_app?
Once again thank you in advance. Any comment or guidance to the concerned documentation would be appreciated. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Fri Jul 14, 2006 5:22 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| Quote: |
1) From the statement:
"4. If you want to load balnce messages between the QMs, put them into an MQ clusters. ..."
given the goal is to provide a single access point for external application to target this Active/Active environment, is there any option apart from MQ clustering? |
I believe what Peter means here is that your gateway qmgr (single point of failure) would be in the HACMP. It would also be either a partial or full repository in the MQCluster. (Probably full repository because of the high availability).
The processing servers need not be in the HACMP but you will need multiple servers on different boxes. They host the physical clustered queues used for processing and the apps for processing.
Now you have a distribution point (gateway qmgr in HACMP) and multiple processing points with load balancing. If any of the processing points fails (QMGR failure/down/ put inhibit etc...) the load gets automatically shifted to the remaining processing points...
Enjoy 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| fahrenheit803 |
| Posted: Wed Aug 16, 2006 9:14 am Post subject: |
|
|
Newbie
Joined: 16 Aug 2006
Posts: 2
|
Dear masters, I have exactly the same situation as TTB, Case B.
Given that
1. svr1_qm and svr2_qm form a cluster, say CLU1, and have a cluster queue clu_q.
2. Application app3 on node svr3, which has svr3_qm, needs to send messages to both app1 and app2.
3. svr3_qm is an outside QM, ie. it does not join CLU1.
I have the following questions.
1. Does svr3_qm need to create channels to and from both svr1_qm and svr2_qm?
2. In the remote queue definition on svr3_qm, which QM(svr1_qm/svr2_qm) should be the parameter of RQMNAME?
3. Follow up Q2, suppose svr1_qm is the parameter of RQMNAME. What if svr1_qm is turned off, will the messages be automatically routed to svr2_qm?
I'm new to MQ clustering, could anyone help? Thanks a lot. |
|
| Back to top |
|
|
| fahrenheit803 |
| Posted: Wed Aug 16, 2006 5:14 pm Post subject: |
|
|
Newbie
Joined: 16 Aug 2006
Posts: 2
|
Dear guys,
A further question, we use MQ v6.
Checked from cluster manual Figure 9 that a gateway QM is in the example.Is it necessary?
If gateway QM is not necessary, can I create two instances of of same queue (cluster queue) on my two clustered QMs?
Thanks. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Wed Aug 16, 2006 8:28 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
You can but you may have to use a cluster alias to load balance correctly.
Enjoy 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Thu Aug 17, 2006 1:40 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
A gateway qmgr is necessary in v5.
A gateway qmgr is not necessary in v6, but you have to configure the workload balancing parameters of the channels and the queues.
_________________
I am *not* the model of the modern major general. |
|
| 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
|
|
|
|
