| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| MQ MB high availability layers |
View previous topic :: View next topic |
| Author |
Message
|
| mjpnet |
 Posted: Fri Jun 01, 2007 1:58 am Post subject: MQ MB high availability layers |
 |
|
Novice
Joined: 09 May 2007
Posts: 15
|
Hi,
I would like to design a MQ MB high availability platform with 3 layers
1º Front End Layer with only MQ to receive message from the clients.
2º Middle End Layer with MQ/MQ to process messages.
3º Layer SQL Data base.
I have some doubts concerning the architecture.
Should I configure the 1º layer with “Webspere MQ cluster” for load balancing proposes, the 2º layer with MSCS (Microsoft) to create queue manager to be cluster resources?
Or should all layers be configure as “Webspere MQ clusters” ?
Could you please send me a site /link/ documentation on possible design solutions architecture to MQ/ MB platform?
Thanks for your help
mjp |
|
| Back to top |
|
 |
| Michael Dag |
| Posted: Fri Jun 01, 2007 2:42 am Post subject: |
|
|
Jedi Knight
Joined: 13 Jun 2002
Posts: 2607
Location: The Netherlands (Amsterdam)
|
and where is your Broker (i assume you mean Message Boker by MB) ?
is it in layer 1 or 2
MQ Clustering is not a HA solution, if on Windows you need to go for MSCS Clustering...
_________________
Michael
MQSystems Facebook page |
|
| Back to top |
|
|
| sebastianhirt |
| Posted: Fri Jun 01, 2007 4:33 am Post subject: |
|
|
Yatiri
Joined: 07 Jun 2004
Posts: 620
Location: Germany
|
| Quote: |
| MQ Clustering is not a HA solution, if on Windows you need to go for MSCS Clustering... |
Even though MQ Clustering can add a lot of value to a high availability solution, using continous operations (which comes with the work load management algorythm).
So you might end up using both. |
|
| Back to top |
|
|
| mjpnet |
| Posted: Fri Jun 01, 2007 5:23 am Post subject: HA + Load balancing |
|
|
Novice
Joined: 09 May 2007
Posts: 15
|
Thanks for your help,
Yes MB is my message broker. And my initial idea was to put it on layer 2.
And I’m note sure if I understand your recommendations.
Should I configure my platform with 2 independent Clusters ? Both with MSCS Cluster + MQ Clustering.
Or should I use MQ Clustering in 1º front end layer to load balancing proposes, And MSCS cluster to 2º layer ?
Or should I use any other platform architecture? |
|
| Back to top |
|
|
| mvenu |
| Posted: Fri Jun 01, 2007 6:57 am Post subject: Re: MQ MB high availability layers |
|
|
Novice
Joined: 06 May 2007
Posts: 10
Location: US
|
Should I configure the 1º layer with “Webspere MQ cluster” for load balancing proposes, the 2º layer with MSCS (Microsoft) to create queue manager to be cluster resources?
===========================================
For high availability & load balancing, you can consider the following however, there are number of possible options:
- Configure Layer 1 MQ as HA with active/standby i.e using HACMP (AIX), MSCS (Microsoft) or Veritas product. This provides high availabity of your entry point.
Configure Layer 2 MQ/MB as HA with mutual take over ie. both are active with mutual failover.
should all layers be configure as “Webspere MQ clusters” ?
=====================================
It depends. Layer 1 MQ & Layer 2 MQ can be in one MQ Cluster or two different MQ Clusters.
Thanks
Venu
_________________
- Venu
- IBM Certified SOA Solution Designer
- IBM Certified System Administrator - WebSphere MQ V6
- IBM Certified Solution Designer - WebSphere MQ V6 |
|
| Back to top |
|
|
| mjpnet |
| Posted: Fri Jun 01, 2007 8:51 am Post subject: HA |
|
|
Novice
Joined: 09 May 2007
Posts: 15
|
I think now I’m starting to see it more clear, although, I have some doubts in what concerns how far we can go.
If in the first layer we use HA with MSCS with 3 active serves and 1 passive, should all the queue manager be in the same share disk?
With this configuration will we have load balancing to all messages that connect to the this front end? All the systems will share the same queue ?
Is it possible to configure the platform to use both MSCS and MQ Cluster technologies at the same time.
Concerning the 2º layer do you mean “use MQ Cluster” technologies? Or also MSCS Active / Active. And in any of the previous solutions we will not use MQ Cluster ?
Do you thing that using “MQ Cluster” just for load balancing in 1 layer and MSCS for Message broker 2 layer, is nor a good or a possible solution ?
Thanks |
|
| Back to top |
|
|
| mvenu |
| Posted: Fri Jun 01, 2007 11:14 am Post subject: Re: HA |
|
|
Novice
Joined: 06 May 2007
Posts: 10
Location: US
|
MQ Cluster & Software clustering products like HACMP, MSCS, Veritas are for different purposes. Since you mentioned MSCS lets discuss with MSCS as example:
Irrespective of the application(if coded for clustering), MSCS provides Failover capability of a application. Usually, an application to failover requires, an ip address, application & data on a shared disk. Upon failover, you need to move your ip, service & shared disk to standby system. These are provided by MSCS.
One of the features of MQ Cluster is load balancing. For example if you have three queue managers QM_A, QM_B, QM_C and you have defined Cluster CLUSQ1 on QM_B & QM_C. if QM_A is your entry point and once a MQ Client put a message CLUSQ1, that will be load balanced between QM_B & QM_C. There are number of options for rounting the messages - Check Queue Manager Clusters Redbook. If QM_C is unavailble, new messages are sent to QM_B. Existing messages on QM_C are marooned until QM_C is recovered.
However, if QM_A is down then its SPOF and there is no way application can send messages to QM_B or QM_C unless they are taken care by the application logic. In order to overcome SPOF of QM_A, you can utilize MSCS with active/standby of QM_A. Similarily you can configure active/active with mutual takeover for Layer 2.
If in the first layer we use HA with MSCS with 3 active serves and 1 passive, should all the queue manager be in the same share disk?
============================================
This is not recommended design. You can configure 3 active and 1 passive with shared disk, but if one queue manager goes down all the Queue Managers need to failover. For the scenerios you described, Its recommended to have each QM a seperate disk & ip.
With this configuration will we have load balancing to all messages that connect to the this front end?
==========================================
Yes, for client connections if you are using an ip sprayer.
All the systems will share the same queue ?
============================
They share definition but not the data. Exception is MQ Z/OS.
_________________
- Venu
- IBM Certified SOA Solution Designer
- IBM Certified System Administrator - WebSphere MQ V6
- IBM Certified Solution Designer - WebSphere MQ V6 |
|
| Back to top |
|
|
| mjpnet |
| Posted: Sun Jun 03, 2007 11:47 am Post subject: Clusters reply |
|
|
Novice
Joined: 09 May 2007
Posts: 15
|
Concerning your last post architecture solution and clarifications, let me say to you that, it is a great solution!! I think we will have a huge amount of advantages with your suggestions.
Please just let me discuss some points.
• Do you think as far as QM_A as my entry messages point, I will not have a bottleneck ? If so how could I avoid it?
• In the 2 layer I want to use 4 MB servers ) why do you do not recommend to use A/A/A/P ?
• A/A/P/P ( Active, Active, Passive , Passive) is it a better solution ? I’m not sure If I understand Why ? could you please help me?
• For the second layer( Message broker) with any of A/A/A/P or A/A/P/P will my queue be all share? And just one for all the brokers?
Thanks a lot once again
Kind regards.
MJP |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Sun Jun 03, 2007 3:05 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Queues are never shared, except on zOS under a queue sharing group on a coupling facility.
You need one and only one MQ cluster for this entire setup.
You should re-read mvenu's messages. There are some very good tips in there, and you should feel extremely grateful that mvenu has decided to do most of your design work for you.
In general, when designing an HA environment, you need to scale each node such that it can handle all work it will ever possibly be asked to carry. This means, for example, if you set up three nodes in an active setup and one in a passive - then all four nodes need to be able to run all three qmgrs simultaneously. If you only scale each node to run one qmgr, then when failover occurs, you've lost capacity. The same with scaling the nodes to only support two of the three qmgrs.
So you need to consider this, yourself. And decide based on your requirements and your budget and the hardware you have available, how many nodes to have active, how many to have passive, and how to configure the failover groups.
And remember the cardinal rule of MQ. No application can ever GET messages from a queue that is not a QLOCAL on the qmgr the application is connected to.
And if you don't have firm requirements, in terms of message throughput and capacity and acceptable downtime... then you need to stop considering HA at all until you get these requirements.
_________________
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
|
|
|
|
