| Author |
Message
|
| SAFraser |
 Posted: Fri Jun 18, 2004 1:51 pm Post subject: (resolved) Solaris/Veritas Cluster failover performance |
 |
|
Shaman
Joined: 22 Oct 2003
Posts: 742
Location: Austin, Texas, USA
|
In browsing posts about Solaris/Veritas clusters, I notice that several members have successfully implemented them.
We are curious about recovery time - from the time of failure, about how long until MQ is fully functional again?
Any other real world observations you have about this redundancy solution will be welcome, too.
Thank you. |
|
| Back to top |
|
 |
| neocruz |
| Posted: Wed Jun 23, 2004 8:13 am Post subject: |
|
|
Acolyte
Joined: 13 Jun 2004
Posts: 54
|
|
| Back to top |
|
|
| RogerLacroix |
| Posted: Wed Jun 23, 2004 9:55 am Post subject: |
|
|
Jedi Knight
Joined: 15 May 2001
Posts: 3264
Location: London, ON Canada
|
Hi,
At my current client site, I setup MQ v5.3 with Veritas on their PROD boxes. The 2 boxes are Sun-Fire-280R (dual CPUs) running MQ in an Active / Passive mode. (Both boxes are running 7/24 but the queue managers are available on 1 box at a time.)
There are 4 active queue managers. Veritas controls the 4 queue managers, Command Servers, listeners and trigger monitors plus the 3 mount points and the VIP (Virtual IP).
To switch (failure-over) MQ from one box to the other, it takes about 4 minutes (some times 5 minutes depending on the load / activity on the queue managers) for the 4 queue managers to be fully active again.
Hope that helps.
Regards,
Roger Lacroix
Capitalware Inc.
_________________
Capitalware: Transforming tomorrow into today.
Connected to MQ!
Twitter |
|
| Back to top |
|
|
| SAFraser |
| Posted: Thu Jun 24, 2004 7:46 am Post subject: |
|
|
Shaman
Joined: 22 Oct 2003
Posts: 742
Location: Austin, Texas, USA
|
Roger, This is exactly the type information I need. We, too, have Sun 280R machines. We will be failing over 6 queue managers and 5 brokers; we estimated 5 minutes in our analysis which sounds like it is pretty close.
If you would permit me one more line of questions.... do you have sender channels? Triggered? Have you been happy with them restarting on their own? How about senders connecting in to your receiver channels-have those restarted clean? Or, perhaps you have client connections coming in on a svrconn? If so, do the apps retry connections for some period of time so that apps do not have to be restarted after failover?
I think our management is very interested in this as a redundancy solution, and it's good to be able to report some real-world experience.
Rich, Thanks for pointer to the supportpac. We have been through it (very useful) as well as through some Veritas info (useful).
Thanks to both of you. |
|
| Back to top |
|
|
| RogerLacroix |
| Posted: Thu Jun 24, 2004 9:00 am Post subject: |
|
|
Jedi Knight
Joined: 15 May 2001
Posts: 3264
Location: London, ON Canada
|
| Quote: |
| do you have sender channels? |
Yes.
Yes.
| Quote: |
| Have you been happy with them restarting on their own? |
Yes. They restart just fine.
| Quote: |
| How about senders connecting in to your receiver channels-have those restarted clean? |
Yes.
| Quote: |
| Or, perhaps you have client connections coming in on a svrconn? |
Yes. 1 queue manager has about 500-600 client channel connections coming from WebLogic applications.
| Quote: |
| If so, do the apps retry connections for some period of time so that apps do not have to be restarted after failover? |
Yes, the WebLogic applications that use MDBs reconnect just fine.
Hope that helps.
Regards,
Roger Lacroix
Capitalware Inc.
_________________
Capitalware: Transforming tomorrow into today.
Connected to MQ!
Twitter |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Thu Jun 24, 2004 10:56 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
| Quote: |
...and 5 brokers; we estimated 5 minutes in our analysis which sounds like it is pretty close.
|
We have Veritas but for Windows. The QMs only take 2 minutes but the brokers take 2-3 times as long. Maybe Solaris will show the same behaviour of the brokers taking longer than the QMs when failing over. Just a heads up. 5 minutes may be a bit optomistic.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| cloud9 |
| Posted: Wed Nov 17, 2004 1:44 pm Post subject: |
|
|
Novice
Joined: 18 Jul 2003
Posts: 13
Location: Jacksonville, FL
|
| What about channel seq number mismatches when the passive queue mgr becomes active ? Is that automatically taken care of because both queue mgrs share the same storage (/var/mqm/qmgrs/.....) ?? |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Wed Nov 17, 2004 3:38 pm Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
Correct. Sequence numbers are maintaned.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| csaunders |
| Posted: Tue Jan 24, 2006 10:00 am Post subject: |
|
|
Acolyte
Joined: 27 Mar 2003
Posts: 64
Location: Arlington, Va
|
| RogerLacroix wrote: |
Hi,
At my current client site, I setup MQ v5.3 with Veritas on their PROD boxes. The 2 boxes are Sun-Fire-280R (dual CPUs) running MQ in an Active / Passive mode. (Both boxes are running 7/24 but the queue managers are available on 1 box at a time.)
There are 4 active queue managers. Veritas controls the 4 queue managers, Command Servers, listeners and trigger monitors plus the 3 mount points and the VIP (Virtual IP).
To switch (failure-over) MQ from one box to the other, it takes about 4 minutes (some times 5 minutes depending on the load / activity on the queue managers) for the 4 queue managers to be fully active again.
Hope that helps.
Regards,
Roger Lacroix
Capitalware Inc. |
roger did you use the support pac from IBM? |
|
| Back to top |
|
|
| alvinlimtc |
| Posted: Sun Jan 29, 2006 12:42 am Post subject: |
|
|
Apprentice
Joined: 22 Dec 2005
Posts: 29
Location: Singapore
|
Just wanted to add that we tried the support pack MC6A for MQ V6 on Solari9 with VCS cluster and it seems to work.
I guess this is because the working directory structure for MQ V6 has not changed from MQ V5.3 |
|
| Back to top |
|
|
| alvinlimtc |
| Posted: Sun Jan 29, 2006 1:13 am Post subject: |
|
|
Apprentice
Joined: 22 Dec 2005
Posts: 29
Location: Singapore
|
Just wanted to add that we tried the support pack MC6A for MQ V6 on Solari9 with VCS cluster and it seems to work.
I guess this is because the working directory structure for MQ V6 has not changed from MQ V5.3 |
|
| Back to top |
|
|
|
|
