| Author |
Message
|
| tychak |
 Posted: Sun Mar 21, 2004 6:04 am Post subject: Cold standby configuration |
 |
|
Novice
Joined: 16 Mar 2004
Posts: 19
|
I am thinking about a cold standby configuration:
For AIX Server 1 and 2, both there is a queue manager with same name, same local queue, remote queue, channels - are all the same. We're going to run HACMP to perform IP failover between the two servers.
So, in normal case, MQ in Server 1 will be running. Another QM at server 3 will communicate with the QM in server 1 by the service IP assigned by HACMP and do the jobs. When server 1 is down, by HACMP, the service IP will be transferred to server 2 and the MQ at server 2 will start and work as if it's server 1.
I discover that the sequence no. used in sender channels has to be reset after switched to another server. So, it wil be done by scripts and called by HACMP during failover.
Can anyone give me some comment on this configuration? Is there any better way to achive this? (MQ Clustering is not preferred as remote servers like server 3 are separated from server 1 & 2 by WAN and it is not desirable to have clustering information exchange through WAN and consumes bandwidth
Thanks very much. |
|
| Back to top |
|
 |
| EddieA |
| Posted: Sun Mar 21, 2004 1:33 pm Post subject: |
|
|
Jedi
Joined: 28 Jun 2001
Posts: 2453
Location: Los Angeles
|
Get the Support Pac on setting up HACMP. (Can't remember the number.  ).
You don't have 2 unique Queue Managers each with their own queue 'storage'. All the data is placed on a shared disk, so when HACMP switches the 'machines', everything still looks the same.
Cheers,
_________________
Eddie Atherton
IBM Certified Solution Developer - WebSphere Message Broker V6.1
IBM Certified Solution Developer - WebSphere Message Broker V7.0 |
|
| Back to top |
|
|
| kirani |
| Posted: Sun Mar 21, 2004 6:21 pm Post subject: |
|
|
Jedi Knight
Joined: 05 Sep 2001
Posts: 3779
Location: Torrance, CA, USA
|
Here is the supportPac information MC63: WebSphere MQ for AIX - Implementing with HACMP
URL: http://www-1.ibm.com/support/docview.wss?rs=203&uid=swg24006416&loc=en_US&cs=utf-8&lang=en
Please note that when the fail-over occurs you will loose non-persistent messages. You might want to look at NPMCLASS queue attribute, which is introduced in CSD6.
_________________
Kiran
IBM Cert. Solution Designer & System Administrator - WBIMB V5
IBM Cert. Solutions Expert - WMQI
IBM Cert. Specialist - WMQI, MQSeries
IBM Cert. Developer - MQSeries
|
|
| Back to top |
|
|
| tychak |
| Posted: Sun Mar 21, 2004 8:20 pm Post subject: |
|
|
Novice
Joined: 16 Mar 2004
Posts: 19
|
If we don't have shared disk for the two servers, is there any way to achieve the repository shared between the two servers?
Thanks. |
|
| Back to top |
|
|
| Stuart_Johnston |
| Posted: Tue Mar 23, 2004 3:47 am Post subject: A thought ! |
|
|
Newbie
Joined: 13 Jan 2003
Posts: 6
|
Greetings,
You will need the Sharded disks for the messages and the queue manager information i.e channel numbers / security and so on...
Best Rgds
Stuart
A possible solution would be to create an MQ Cluster.
This would mean that you still would have connectivity and message flow
BUT the messages on the queue manager which erred for which ever reason wouldn't be available until it rejoins. |
|
| Back to top |
|
|
| tychak |
| Posted: Tue Mar 23, 2004 6:19 pm Post subject: |
|
|
Novice
Joined: 16 Mar 2004
Posts: 19
|
Thanks for your advise.
But from my first post, my aim is to setup two queue managers with same configuration to achieve cold standby purpose. I know that it will be preferred to use HACMP and share the repository by shared disk. However, due to lack of budget, no shared disk can be brought. Therefore, is it possible achive my goal if the two repository of the two queue managers are not the same?
Thanks. |
|
| Back to top |
|
|
| offshore |
| Posted: Wed Mar 24, 2004 4:23 am Post subject: |
|
|
Master
Joined: 20 Jun 2002
Posts: 222
|
Is this queue manager part of a cluster? By cold standby, do you mean another QMGR and server box (same ip, same hostname, ect) turned off waiting to be used in case of emergency?
If the QMGR is not part of a a cluster then I say yes it is possible. If that QMGR (or more likely) the server goes toes up, then just power the other one on and away you go. *Note:You'll have to reset the channel seq numbers* Obviously, there is the chance that some data could be lost since your backup QMGR has no idea what your original QMGR was doing at the time something went wrong.
If the QMGR is part of a cluster w/o HACMP and shared disk, then I would say no. Once you try to bring up your backup QMGR and let it join the cluster the QMID will be different, the repository is going to be confused, ect.
Its possible that somehow you could get it to work, but I think you're asking for a lot of problems and troubles.
Just my $0.00002 |
|
| Back to top |
|
|
| tychak |
| Posted: Wed Mar 24, 2004 6:10 pm Post subject: |
|
|
Novice
Joined: 16 Mar 2004
Posts: 19
|
| Thanks for your reply. The former case you mentioned is that what I want. |
|
| Back to top |
|
|
|
|
