| Author |
Message
|
| echoesian |
 Posted: Sun Oct 05, 2008 7:29 pm Post subject: Swing in Cluster |
 |
|
Apprentice
Joined: 30 May 2008
Posts: 33
|
I have a doubt here I cannot confirm this will work, hope masters can help me:
Cluster: EAICLUSTER
1) MQ1 (FR)
2) MQ2 (FR) OFFLINE
3) HOST (FR)
Active
MQ1 -> HOST
Changing to
MQ1 --X-- HOST
MQ2 ----> HOST
I want to use MQ2 as a backup queue manager to connect to Host. When MQ1 is down, I will bring up the MQ2 and connect to the Host. I know the things that need to change such as connection names but I do not know any implication of it:
1) Shutdown Cluster sender on MQ1
2) Change cluster sender in HOST to point to MQ2
2) Change connection name of cluster sender in MQ2 to point to the HOST
3) Reset channel or maybe resolve
4) Reset EAICLUSTER.
Does the above steps sufficient in order to run? |
|
| Back to top |
|
 |
| Mr Butcher |
| Posted: Mon Oct 06, 2008 12:15 am Post subject: |
|
|
Padawan
Joined: 23 May 2005
Posts: 1716
|
if this is a cluster (a MQ Cluster), why not leave all channels as they are? in a MQ cluster, i do not see the need to change connection names in cluster channels when one queuemanager is stopped and another one is started.
i also do not see the need for channel or cluster reset commands.
what you describe is required when using a cold standby queuemanager without clustering.
_________________
Regards, Butcher |
|
| Back to top |
|
|
| echoesian |
| Posted: Mon Oct 06, 2008 12:34 am Post subject: |
|
|
Apprentice
Joined: 30 May 2008
Posts: 33
|
| Mr Butcher wrote: |
if this is a cluster (a MQ Cluster), why not leave all channels as they are? in a MQ cluster, i do not see the need to change connection names in cluster channels when one queuemanager is stopped and another one is started.
i also do not see the need for channel or cluster reset commands.
what you describe is required when using a cold standby queuemanager without clustering. |
Thanks for reply. No, The MQ2 is a backup node which is always offline. It will bring up when MQ1 (Production) down. This is because it is located at a remote DR site which has its own EAI environment connecting to a DR Host.
Yes, it is something like a cold standby but using the DR server as a backup instead. The DR is being setup to be able to connect to the DR Host and DR front-end systems.
After the MQ1 (Production) is up, the MQ2 needs to be moved back to the DR setup. Everything need to be reconfigured again.
I know this complicate things and involves very high risks, but the no choice the client wants it.  |
|
| Back to top |
|
|
| Mr Butcher |
| Posted: Mon Oct 06, 2008 12:48 am Post subject: |
|
|
Padawan
Joined: 23 May 2005
Posts: 1716
|
yes, the client....
but why do you use mq clustering then?
if you have mq2 as a cold standby, and mq1 fails, then (assuming the channel host to mqx and mqx to host are the same names on mq1 and mq2 queuemanagers)
* channel host.to.mqx has to be stopped, conname changed to point to mq2, reset, started
* channel mqx to host (from mq2) has to be reset and started
now why use MQ Cluster?
_________________
Regards, Butcher |
|
| Back to top |
|
|
| zhanghz |
| Posted: Mon Oct 06, 2008 1:48 am Post subject: |
|
|
Disciple
Joined: 17 Jun 2008
Posts: 186
|
to me, DR server should be used when there is a disaster at production server. If the production server is only temporarilly down and can be brought up in, say, 1 or 2 hours, better off trying to bring up the production server asap.
to use DR server, you need to make sure the DR qmgr has the most updated mq objects and persistent messages. If MQ clients are used, should anything be done on client side (i am not sure about client)?
As Mr Butcher pointed, why use cluster if there are only 2 qmgrs here?
If cluster is used, a refresh cluster might be needed.
If hardware HA (high availability) can be done and if MQ can be set up on a HA machine, I guess better use HA for failover? |
|
| Back to top |
|
|
| echoesian |
| Posted: Mon Oct 06, 2008 5:00 am Post subject: |
|
|
Apprentice
Joined: 30 May 2008
Posts: 33
|
| Mr Butcher wrote: |
yes, the client....
but why do you use mq clustering then?
if you have mq2 as a cold standby, and mq1 fails, then (assuming the channel host to mqx and mqx to host are the same names on mq1 and mq2 queuemanagers)
* channel host.to.mqx has to be stopped, conname changed to point to mq2, reset, started
* channel mqx to host (from mq2) has to be reset and started
now why use MQ Cluster? |
Ok.... the cluster is used for the two Brokers workload balancing for Production.
Production environment has:
EAICLUSTER:
APPS, BROKER1 (MQ1), BROKER2 (MQ2), HOST |
|
| Back to top |
|
|
| echoesian |
| Posted: Mon Oct 06, 2008 5:10 am Post subject: |
|
|
Apprentice
Joined: 30 May 2008
Posts: 33
|
| zhanghz wrote: |
to me, DR server should be used when there is a disaster at production server. If the production server is only temporarilly down and can be brought up in, say, 1 or 2 hours, better off trying to bring up the production server asap.
to use DR server, you need to make sure the DR qmgr has the most updated mq objects and persistent messages. If MQ clients are used, should anything be done on client side (i am not sure about client)?
As Mr Butcher pointed, why use cluster if there are only 2 qmgrs here?
If cluster is used, a refresh cluster might be needed.
If hardware HA (high availability) can be done and if MQ can be set up on a HA machine, I guess better use HA for failover? |
I also feel it is very messy to do like this but the reason being the bank has a very bad experience on the HA failover, there was a case the failover could not be executed properly and the bank suffered nearly half day downtime! |
|
| Back to top |
|
|
| Mr Butcher |
| Posted: Mon Oct 06, 2008 5:23 am Post subject: |
|
|
Padawan
Joined: 23 May 2005
Posts: 1716
|
| Quote: |
Production environment has:
EAICLUSTER:
APPS, BROKER1 (MQ1), BROKER2 (MQ2), HOST |
how that? above you wrote, MQ2 is offline, just started for backup purposes when it is needed, but now you write it is in a cluster with MQ1 and has a broker for workload balancing?
_________________
Regards, Butcher |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Mon Oct 06, 2008 6:39 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
| echoesian wrote: |
| I also feel it is very messy to do like this but the reason being the bank has a very bad experience on the HA failover, there was a case the failover could not be executed properly and the bank suffered nearly half day downtime! |
Just because the correct solution wasn't implemented correctly doesn't mean you abandon it and cobble together something less appropriate.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| echoesian |
| Posted: Mon Oct 06, 2008 6:45 am Post subject: |
|
|
Apprentice
Joined: 30 May 2008
Posts: 33
|
| Mr Butcher wrote: |
| Quote: |
Production environment has:
EAICLUSTER:
APPS, BROKER1 (MQ1), BROKER2 (MQ2), HOST |
how that? above you wrote, MQ2 is offline, just started for backup purposes when it is needed, but now you write it is in a cluster with MQ1 and has a broker for workload balancing? |
Just for example, ok I re-write my statement again to avoid confusion...
Production Environment
EAICLUSTER:
PROD APPS, PROD BROKER1 (MQ1), PROD BROKER2 (MQ3), PROD HOST
DR Environment (OFFLINE)
EAICLUSTER:
DR APPS, DR BROKER (MQ2), DR HOST
[/i] |
|
| Back to top |
|
|
| Mr Butcher |
| Posted: Mon Oct 06, 2008 7:09 am Post subject: |
|
|
Padawan
Joined: 23 May 2005
Posts: 1716
|
if all these queuemanagers are in a cluster, i do not understand why you want to do things like conname change in the cluster channels to make mq2 talk to host. it is just a matter of the cluster setup for me (e.g. pick available target queue)
_________________
Regards, Butcher |
|
| Back to top |
|
|
| echoesian |
| Posted: Mon Oct 06, 2008 6:21 pm Post subject: |
|
|
Apprentice
Joined: 30 May 2008
Posts: 33
|
| Mr Butcher wrote: |
| if all these queuemanagers are in a cluster, i do not understand why you want to do things like conname change in the cluster channels to make mq2 talk to host. it is just a matter of the cluster setup for me (e.g. pick available target queue) |
No, the DR cluster is always offline. It will only bring up during DR on Production. |
|
| Back to top |
|
|
| Mr Butcher |
| Posted: Mon Oct 06, 2008 10:25 pm Post subject: |
|
|
Padawan
Joined: 23 May 2005
Posts: 1716
|
again - even if it is offline and only used in case of desaster, the queuemanager is member of a cluster. so it has cluster sender and cluster receiver. if you start it, it will join the cluster. maybe you need to refresh because it has been offline for a too long period of time. but why do you think you have to alter channels, connames, .... ?!? thats not the concept of a mq cluster.
_________________
Regards, Butcher |
|
| Back to top |
|
|
| exerk |
| Posted: Tue Oct 07, 2008 1:16 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
And to have a Full Repository off-line indefinitely (how often is DR invoked? Hopefully never) seems like a bad idea anyway, irrespective of having three (see Mr. Potkay's very illuminating post somewhere in this forum).
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
|
|
