| Author |
Message
|
| memphis |
 Posted: Sun Jan 17, 2010 10:26 pm Post subject: HA test in MQ 7.0.1 using contact admin |
 |
|
Novice
Joined: 05 Jan 2010
Posts: 21
|
Hi All,
I was able to succesfully test the MQ HA feature in 7.0.1 using contact admin. I used 3 VM images of openSuse 11.2 for my test (one of the image is solely used for storage purpose).
I was succesfully able to transfer the control to the standby server from the active server when issued endmqm -s QMGRNAME command (active server) and also tested succesfully when the server is down.
I want to test various possibilities of the HA configuration and, hence, I killed the processes of mqm (all processess at the same time) but this doesn't seem to work here.
I believe that a true HA solution should also accomodate such a criteria, isn't it? Can someone shed some light on why this won't work? (or) Did I miss something here..
I appreciate your replies..
Thank you |
|
| Back to top |
|
 |
| fjb_saper |
| Posted: Mon Jan 18, 2010 7:57 am Post subject: Re: HA test in MQ 7.0.1 using contact admin |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| memphis wrote: |
Hi All,
I want to test various possibilities of the HA configuration and, hence, I killed the processes of mqm (all processess at the same time) but this doesn't seem to work here.
I believe that a true HA solution should also accomodate such a criteria, isn't it? Can someone shed some light on why this won't work? (or) Did I miss something here..
I appreciate your replies..
Thank you |
Did you kill the processes in the order specified by the admin manual? 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon Jan 18, 2010 8:45 am Post subject: Re: HA test in MQ 7.0.1 using contact admin |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| fjb_saper wrote: |
| Did you kill the processes in the order specified by the admin manual? |
I think in the context of this test it's valid to kill the processes in a non-documented way. In the event of a server crash it's unlikely the queue manager will go down neatly.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Mon Jan 18, 2010 7:09 pm Post subject: Re: HA test in MQ 7.0.1 using contact admin |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| Vitor wrote: |
| fjb_saper wrote: |
| Did you kill the processes in the order specified by the admin manual? |
I think in the context of this test it's valid to kill the processes in a non-documented way. In the event of a server crash it's unlikely the queue manager will go down neatly. |
Evidently not ok to kill processes in a non documented way.
If you use HA software do you expect the failover to come up if you killed the HA software first? Remember here MQ plays as well the role of HA software...
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon Jan 18, 2010 7:26 pm Post subject: Re: HA test in MQ 7.0.1 using contact admin |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| fjb_saper wrote: |
| Remember here MQ plays as well the role of HA software... |
I can't remember what I didn't realise in the first place! I am enriched with new knowledge 
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon Jan 18, 2010 7:31 pm Post subject: Re: HA test in MQ 7.0.1 using contact admin |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| memphis wrote: |
I believe that a true HA solution should also accomodate such a criteria, isn't it? Can someone shed some light on why this won't work? (or) Did I miss something here..
|
I think what we both missed here is that the new feature isn't intended to replace HA software for WMQ, but fill a gap in the functionality (I've been doing a little supplimentary reading). Hence you're right; a true HA solution would accomodate this scenario. But this isn't a true HA solution.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| memphis |
| Posted: Mon Jan 18, 2010 7:31 pm Post subject: Results of my test |
|
|
Novice
Joined: 05 Jan 2010
Posts: 21
|
I succcesfully demonstrated reconnecting of the Clients by the following methods:
1. Use endmqm -s QMGR (switchover)
2. Shutdown of the server running active QMGR (failover)
3. Killing all mqm processes at once on the machine running active QMGR instance (failover)
But I am not able to demonstrate the ability to failover when the machine is POWERED-OFF. I'm very curious to why it is not failing over when I pull the plug off of the machine running active instance of the QMGR.
Anybody have ideas on this? Appreciate your replies gentlemen... |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Mon Jan 18, 2010 8:45 pm Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
| But I am not able to demonstrate the ability to failover when the machine is POWERED-OFF. |
I'd rather not guess at what you mean by this. A bit more detail, please. What did happen when you powered the box off? Anything?
Did the disk drives get powered off, too?
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| memphis |
| Posted: Mon Jan 18, 2010 10:14 pm Post subject: Scenario goes like this |
|
|
Novice
Joined: 05 Jan 2010
Posts: 21
|
I'm testing my current HA configuration (contact admin) in 7.0.1 in 3 virtual machines (VMs) created by me for that purpose.
I installed MQ 7.0.1 server software on 2 of the VMs and the third VM is used solely for the contact admin (storage of MQ logs and qmgr data).
I have mounted the network drive onto the 2 VMs (hosting active and standby queue managers)
Now, I want to test this HA configuration by powering off one of the VMs of the 2 (which is hosting the active queue manager) and see if it would failover to the other server. (I can power off my VM without shutting it down)
Wanted to check out this possibility and all my attempts to check this failed.
Will such a failover possible in this case?
Please share your thoughts.. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Mon Jan 18, 2010 10:39 pm Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
| Now, I want to test this HA configuration by powering off one of the VMs of the 2 (which is hosting the active queue manager) and see if it would failover to the other server. |
Have you tried it?
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| memphis |
| Posted: Mon Jan 18, 2010 10:45 pm Post subject: |
|
|
Novice
Joined: 05 Jan 2010
Posts: 21
|
Yes Sir.. I tried it many a times but my efforts were in vain.
The sample program just tries reconnecting without being failing over to the standby machine.
Whereas, all other options are working fine.. Shutting down the VM (with active instance), killing the mqm processes on the server (with active instance), issuing endmqm -s etc.,
But powering off the VM is not working..
So, I got another question in my mind... I guess it has something to do with powering off the VM. Shutting down a VM probably sends out a signal that the machine is going to be shutdown and mq picks it up (i guess) whereas Powering-off VM is a sudden shutdown without any intimation..
Can this be an issue for the clients not being able to reconnect? Please share your thoughts on this!!!!
Thank you |
|
| Back to top |
|
|
| mvic |
| Posted: Tue Jan 19, 2010 1:17 am Post subject: Re: Results of my test |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| memphis wrote: |
I succcesfully demonstrated reconnecting of the Clients by the following methods:
1. Use endmqm -s QMGR (switchover)
2. Shutdown of the server running active QMGR (failover)
3. Killing all mqm processes at once on the machine running active QMGR instance (failover)
|
This is different from your original post. Your original post said
| Quote: |
| I killed the processes of mqm (all processess at the same time) but this doesn't seem to work here. |
Did something change to make this work?
| Quote: |
| But I am not able to demonstrate the ability to failover when the machine is POWERED-OFF. I'm very curious to why it is not failing over when I pull the plug off of the machine running active instance of the QMGR. |
Like bruce I wonder what you mean here. How do you power off a VM, and how does this differ from "Shutdown of the server running active QMGR", point 2 in your list? |
|
| Back to top |
|
|
| mvic |
| Posted: Tue Jan 19, 2010 1:20 am Post subject: |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| memphis wrote: |
| The sample program just tries reconnecting without being failing over to the standby machine. |
Assuming this is a reconnectable client...
It will take some time (not sure if this time is defined anywhere) for the connect attempt to fail. Then if the client has a second IP address to use, it will use it. Did you wait long enough for the program to try the second IP address? |
|
| Back to top |
|
|
| memphis |
| Posted: Tue Jan 19, 2010 1:29 am Post subject: |
|
|
Novice
Joined: 05 Jan 2010
Posts: 21
|
Yes, I totally understand that it takes some time to reconnect. I waited for as long as 1.5 hours with absolutely no use.
The client automatically gets disconnected after certain period of time.
Even after so much of time, the client DOES NOT reconnect to the other server and also the standy QMGR (which I expect to go active) still remains in the standby state.
Any further advise is greatly appreciated.. |
|
| Back to top |
|
|
| exerk |
| Posted: Tue Jan 19, 2010 1:41 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
I had something similar to memphis happen when I was also testing using VM's; in that case I was using the amqsghac, amqsphac and amqsmhac samples. Unfortunately, I can't bring anything more to the party as I did not have time to follow up on the possible causes, or try it on hardware to determine whether it was a VM issue, or configuration.
_________________
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 |
|
|
|
|
