Author |
Message
|
ankurlodhi |
Posted: Thu Mar 22, 2018 10:35 am Post subject: Converting MQ standalone to Microsoft CLuster |
|
|
Master
Joined: 19 Oct 2010 Posts: 266
|
We have a standalone MQ Development server, Business wants to move some more queue managers on this box and same in production. But they want to make these queue manager HA using mscs in production but don't want to do the same in Development server. is it ok not test these Queue managers in HA Development cluster. |
|
Back to top |
|
 |
Vitor |
Posted: Thu Mar 22, 2018 11:38 am Post subject: |
|
|
 Grand High Poobah
Joined: 11 Nov 2005 Posts: 26093 Location: Texas, USA
|
Are you asking if it's ok to not have HA in Dev, if it's ok to set up HA in Production without testing MCSC in Dev or if an MQ program developed on a stand alone queue manager will work on a queue manager clustered in MCSC?
Between your English and my English your post isn't working for me
So:
- You can put any queue manager under MCSC, Dev or Prod
- Your application shouldn't know the queue manager it's using is under MCSC
- MCSC works as well and is as easy to configure as any MSoft product i.e. it's like juggling chainsaws - possible but requires practice and can get messy
- From that, you can put your Prod queue managers under MCSC without testing it in Dev first. You just have to accept that there's a risk of it getting messy
- if you put Dev queue managers under HA, you get more uptime for your developers but no business benefit long term
- if you put Dev queue managers under HA and they're hosed, you get less business impact short term
If none of those answers where in fact answers to your question, accept my apologies, pretend I'm an MCSC cluster and have another go. _________________ Honesty is the best policy.
Insanity is the best defence. |
|
Back to top |
|
 |
ankurlodhi |
Posted: Thu Mar 22, 2018 12:18 pm Post subject: |
|
|
Master
Joined: 19 Oct 2010 Posts: 266
|
Yes, I want to know if it ok to not have a MCSC in Dev. |
|
Back to top |
|
 |
Vitor |
Posted: Thu Mar 22, 2018 12:25 pm Post subject: |
|
|
 Grand High Poobah
Joined: 11 Nov 2005 Posts: 26093 Location: Texas, USA
|
ankurlodhi wrote: |
Yes, I want to know if it ok to not have a MCSC in Dev. |
Then I think I've answered your questions, albeit mostly by accident.
Yay me.  _________________ Honesty is the best policy.
Insanity is the best defence. |
|
Back to top |
|
 |
ankurlodhi |
Posted: Thu Mar 22, 2018 12:54 pm Post subject: |
|
|
Master
Joined: 19 Oct 2010 Posts: 266
|
as always you're great. but now the real issue what i am concerned about it.
these Queue managers use many local service accounts and local groups and some are part of MQM group as . won't there be issues due to these groups as there GID/UID might not be same on node1 and node 2.
I want to move Dev to HA cluster only for testing purpose of these issues but my management doesn't understand this. |
|
Back to top |
|
 |
fjb_saper |
Posted: Thu Mar 22, 2018 6:20 pm Post subject: |
|
|
 Grand High Poobah
Joined: 18 Nov 2003 Posts: 20756 Location: LI,NY
|
ankurlodhi wrote: |
as always you're great. but now the real issue what i am concerned about it.
these Queue managers use many local service accounts and local groups and some are part of MQM group as . won't there be issues due to these groups as there GID/UID might not be same on node1 and node 2.
I want to move Dev to HA cluster only for testing purpose of these issues but my management doesn't understand this. |
No that's why your groups and ids are all in active directory...  _________________ MQ & Broker admin |
|
Back to top |
|
 |
PeterPotkay |
Posted: Sun Mar 25, 2018 1:48 pm Post subject: |
|
|
 Poobah
Joined: 15 May 2001 Posts: 7722
|
There was an APAR a few years back where there where one of the Fixpacks introduced an issue with a failover for a queue manager under MSCS control.
If the only place you have a MSCS cluster is in Production, then that's the first place you find out there are problems in this space. Production is the first place you get to experiment with and figure out procedures for a queue manager under MSCS control?
Having a unique configuration in Production is brave. Or foolish. Depends on your perspective I suppose. _________________ Peter Potkay
Keep Calm and MQ On |
|
Back to top |
|
 |
gbaddeley |
Posted: Sun Mar 25, 2018 9:44 pm Post subject: |
|
|
 Jedi Knight
Joined: 25 Mar 2003 Posts: 2538 Location: Melbourne, Australia
|
PeterPotkay wrote: |
There was an APAR a few years back where there where one of the Fixpacks introduced an issue with a failover for a queue manager under MSCS control.
If the only place you have a MSCS cluster is in Production, then that's the first place you find out there are problems in this space. Production is the first place you get to experiment with and figure out procedures for a queue manager under MSCS control?
Having a unique configuration in Production is brave. Or foolish. Depends on your perspective I suppose. |
We have faced the same issue, where prod is the only MSCS clustered environment.
Many of our apps have a QA, Regression, Volume Test or Performance environment (call it what you will) which is configured and sized like Production. In some cases the app can't afford to do this, so they need to accept the risk that they can't exercise the setup / configuration process and possible failure modes in the lower environments. _________________ Glenn |
|
Back to top |
|
 |
ankurlodhi |
Posted: Mon Mar 26, 2018 3:31 pm Post subject: |
|
|
Master
Joined: 19 Oct 2010 Posts: 266
|
fjb_saper wrote: |
No that's why your groups and ids are all in active directory...  |
that's the issue many groups are not on active directory they are local that's the entire reason i want to test thing in dev before jumping into prod. |
|
Back to top |
|
 |
fjb_saper |
Posted: Mon Mar 26, 2018 8:19 pm Post subject: |
|
|
 Grand High Poobah
Joined: 18 Nov 2003 Posts: 20756 Location: LI,NY
|
ankurlodhi wrote: |
fjb_saper wrote: |
No that's why your groups and ids are all in active directory...  |
that's the issue many groups are not on active directory they are local that's the entire reason i want to test thing in dev before jumping into prod. |
That's a whole world of hurt just right off the bat.
As you groups are local they are not going to have the same SID/GID on the other nodes of the cluster, and you will always have to create them and authorize them twice...  _________________ MQ & Broker admin |
|
Back to top |
|
 |
mvic |
Posted: Thu Mar 29, 2018 2:49 pm Post subject: |
|
|
 Jedi
Joined: 09 Mar 2004 Posts: 2080
|
ankurlodhi wrote: |
I want to move Dev to HA cluster only for testing purpose of these issues but my management doesn't understand this. |
Similar to what's been said, but this is my 2p worth.
Mention these scenarios to your management:
You have an in-production MSCS cluster.
Scenario 1: You want to develop new application features and test that they work ok when using MSCS for node-to-node failover or whatever, and you want to do this while the production servers continue working.
Scenario 2: Or, you want to replicate a failover problem in a non-production environment to debug it, and again you want to do this while the production servers continue working.
These are the typical reasons why people buy more servers than just the ones they want for production.
I don't know but perhaps your site is willing to accept risks, in which case I refer you to the remarks of the most learned Vitor (with emphasis added by me):
Quote: |
you can put your Prod queue managers under MCSC without testing it in Dev first. You just have to accept that there's a risk of it getting messy |
And the most learned PeterPotkay:
Quote: |
Having a unique configuration in Production is brave. Or foolish. Depends on your perspective I suppose. |
|
|
Back to top |
|
 |
PeterPotkay |
Posted: Mon Apr 02, 2018 5:17 pm Post subject: |
|
|
 Poobah
Joined: 15 May 2001 Posts: 7722
|
PeterPotkay wrote: |
There was an APAR a few years back where there where one of the Fixpacks introduced an issue with a failover for a queue manager under MSCS control.
If the only place you have a MSCS cluster is in Production, then that's the first place you find out there are problems in this space. Production is the first place you get to experiment with and figure out procedures for a queue manager under MSCS control?
|
Here is the (old) APAR that prevented a queue manager from coming up on the passive node in the MSCS cluster.
http://www-01.ibm.com/support/docview.wss?uid=swg1IC74752
Imagine our surprise/frustration when we applied Fix Pack 7.0.1.4 to a functional MSCS cluster, failed the node over to patch the primary, and the queue manager would not start. Bad enough we ran into this in the Lab, luckily we had a 2 node play cluster.
But imagine finding that out for the first time in Production at 2 AM!  _________________ Peter Potkay
Keep Calm and MQ On
Last edited by PeterPotkay on Mon Apr 02, 2018 7:11 pm; edited 1 time in total |
|
Back to top |
|
 |
Vitor |
Posted: Mon Apr 02, 2018 6:28 pm Post subject: |
|
|
 Grand High Poobah
Joined: 11 Nov 2005 Posts: 26093 Location: Texas, USA
|
fjb_saper wrote: |
That's a whole world of hurt just right off the bat. |
One of the chain saws just fell on your arm.
fjb_saper wrote: |
As you groups are local they are not going to have the same SID/GID on the other nodes of the cluster, and you will always have to create them and authorize them twice...  |
And because the local SID/GID combinations will be different in Dev & Prod, you'll never be sure it works until you try it (with due deference to @PeterPotkay) at 2am one dark and stormy night. _________________ Honesty is the best policy.
Insanity is the best defence. |
|
Back to top |
|
 |
fjb_saper |
Posted: Tue Apr 03, 2018 8:41 pm Post subject: |
|
|
 Grand High Poobah
Joined: 18 Nov 2003 Posts: 20756 Location: LI,NY
|
Vitor wrote: |
One of the chain saws just fell on your arm.
fjb_saper wrote: |
As you groups are local they are not going to have the same SID/GID on the other nodes of the cluster, and you will always have to create them and authorize them twice...  |
And because the local SID/GID combinations will be different in Dev & Prod, you'll never be sure it works until you try it (with due deference to @PeterPotkay) at 2am one dark and stormy night. |
And because this is a dark and stormy night there is no help available from the AD Team and meanwhile production is down and when it comes to root cause analysis it's all your fault!!!  _________________ MQ & Broker admin |
|
Back to top |
|
 |
|