| Author |
Message
|
| George Carey |
 Posted: Tue Sep 15, 2009 8:52 am Post subject: modern NFS file systems |
 |
|
Knight
Joined: 29 Jan 2007
Posts: 500
Location: DC
|
Qmgr instances:
| Quote: |
"With this new availability option, the messages and data for a multi-instance queue manager are held on networked storage accessed via a modern network file system protocol, such as Network File System (NFS) version 4.
Multiple instances of this queue manager can then be defined and started on different machines, with one active instance and one standby instance. The active queue manager instance processes messages and accepts connections from applications and other queue managers. |
Does this quote mean IBM formally endorses use of contact admin network storage for MQ and HA and not just SAN?
(tangent non-sequitur: markt your klein bottle icon took me back to my college days in topology class. Here is a neat 'youtube' of a klein bottle animation showing it as a topological extension of a Mobius strip!)
http://www.youtube.com/watch?v=E8rifKlq5hc
_________________
"Truth is ... grasping the virtually unconditioned",
Bernard F. Lonergan S.J.
(from book titled "Insight" subtitled "A Study of Human Understanding") |
|
| Back to top |
|
 |
| George Carey |
| Posted: Thu Sep 24, 2009 8:43 am Post subject: no takers ?? |
|
|
Knight
Joined: 29 Jan 2007
Posts: 500
Location: DC
|
No takers on this question ???
This is a big deal if it is so !!!
contact admin supported wholely by IBM, even preferred for HA in 7.0.1 + versions, it sounds like.
Hitherto this ... SAN was only formally endorsed network storage architecture supported for WSMQ to my knowledge.
Thought there would be some interest ...will try MQ listers.
_________________
"Truth is ... grasping the virtually unconditioned",
Bernard F. Lonergan S.J.
(from book titled "Insight" subtitled "A Study of Human Understanding") |
|
| Back to top |
|
|
| zpat |
| Posted: Mon Jul 19, 2010 6:12 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
| markt wrote: |
| Quote: |
| I wonder if the client auto-reconnect feature can be implemented without making any application coding changes |
No code changes are required, but you do have to upgrade to use the 701 libs. And have the channels use SHARECNV >0.
To enable the function with no code changes, put the right entry in mqclient.ini. Alternatively, it can be enabled by a minor code change - set the appropriate MQCNO option on MQCONNX. |
Can someone show me an example of mqclient.ini?
Does SHARECNV go in the V7 CCDT?
Does client auto-reconnect work with Java/JMS applications? |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Mon Jul 19, 2010 10:03 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
|
| Back to top |
|
|
| Nick Lethbridge |
| Posted: Mon Nov 22, 2010 9:19 am Post subject: |
|
|
Voyager
Joined: 13 Aug 2001
Posts: 88
Location: Santander, UK
|
I've had a look at the 'Testing and support statement for WMQ multi-instance queue managers' (http://www-01.ibm.com/support/docview.wss?uid=swg21433474).
It mentions that NFS v3 cannot be used and then says the following for Solaris:
Multi-instance queue managers on Solaris using NFS v4 are NOT supported due to a suspected problem in the NFS v4 client. The effect is data corruption of the queue manager logs following a network outage. The issue is being investigated by Oracle. The case reference is CR 6942081.
Am I correct in thinking that WMQ multi-instance queue managers CANNOT be implemented on Sun Solaris?
Or have Oracle provided a fix for the NFS v4 client (since the above statement was issued in June 2010)?
Has anyone managed to get multi-instance queue managers configured and working (with proven fault tolerance) on Sun Solaris?
Regards,
Nick. |
|
| Back to top |
|
|
| CHAFF |
| Posted: Mon Mar 07, 2011 9:48 am Post subject: My .02 cents |
|
|
Newbie
Joined: 07 Mar 2011
Posts: 5
|
So back to the original question of Multi-Instance queues on Solaris.
Ok I've been working on this damn problem for the last 4 months. My final conclusion(for now-it changes daily) is that:
Yes it works. But not officially supported for now (scary)
To do multi-instance queues you require:
-Install 7.0.1.x (use the latest fixpack as there were Multi-Instance Queue fixes in FP3 and AMQMFSCK fixes in FP4)
-Another Solaris server, is used as shared storage.
-(not a must but probably recommended to use) Modified mount:
[IP of contact admin]:/[sharename] - /[local mountpoint] nfs - yes rw,hard,intr,noac,forcedirectio,vers=4
note: Couldn't not make Solaris+NFSv4+contact admin appliances work so could not test that configuration
This is everything I've gathered. Please correct me if I'm wrong because I would gladly receive any additional info. I will update this along with a build notes at some point in time.
references to follow |
|
| Back to top |
|
|
| CHAFF |
| Posted: Mon Mar 07, 2011 9:50 am Post subject: |
|
|
Newbie
Joined: 07 Mar 2011
Posts: 5
|
|
| Back to top |
|
|
| Nick Lethbridge |
| Posted: Wed Mar 09, 2011 3:04 am Post subject: |
|
|
Voyager
Joined: 13 Aug 2001
Posts: 88
Location: Santander, UK
|
Thanks Chaff for your feedback.
It does sound a bit dire though...I remember IBM 'pulling' support for Multi-Instance queue managers on Solaris shortly after V7 was released.
I'd assumed that by now it would be ok, but from what you say it still sounds a combination of FixPacks, interim fixes and crossed fingers....
I'd planned to try this out on a new 2-node UAT cluster, but I've now discovered that we will only get a '1-node UAT cluster' (also known as 'a server'....), so I think I will stick with the current HA/VCS setup in PROD.
Multi-Instance seems to offer a 'faster failover' HA solution (compared to the VCS approach), but it doesn't sound like it's 'production ready' on Sun Solaris? |
|
| Back to top |
|
|
| CHAFF |
| Posted: Wed Mar 09, 2011 8:37 am Post subject: |
|
|
Newbie
Joined: 07 Mar 2011
Posts: 5
|
Hi Nick,
Well the MQ code sounds like its ok. The root cause seems to lie in the Solaris NFSv4 client. Oddly enough during testing I never really found any problems; probably cause I had 2 virtual Solaris hosts + 3rd virtual Solaris host (acting as shared disk) and I never saw issues. However, during testing I also had also experimented with a virtual contact admin appliance and I had issues.
One of those tips sure sounds like the issue only occurs when accessing a real contact admin device which would coincide with what I've experienced thus far. Of course a real NFSv4 contact admin is what would be in prod...so I would have to say yes; its not production ready. But kinda fun to test with otherwise  The failover is very cool, and if you poweroff one of the hosts, the other one takes over/locks the shared disk and and restarts the QM. |
|
| Back to top |
|
|
| indian82 |
| Posted: Thu Mar 31, 2011 10:13 am Post subject: Multi instance on solaris |
|
|
Newbie
Joined: 28 Mar 2011
Posts: 4
|
Hi everubody,
Thanks evrybody for the things you put on this thread...really got lot of information.But just want to know has the problem with NSF V4 has been debugged and did anybody do mq v7 multi instance + solaris + contact admin/NSF V4 and were succesfull and went to production.Thank you guys. |
|
| Back to top |
|
|
| mvic |
| Posted: Thu Mar 31, 2011 11:05 am Post subject: |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| Nick Lethbridge wrote: |
| I remember IBM 'pulling' support for Multi-Instance queue managers on Solaris shortly after V7 was released. |
MI qmgrs were first available in 7.0.1. I seem to remember the problem on Solaris was known, and published. The ibm.com link I have for this is now dead, so the page has been removed - presumably superseded by the link given above ( http://www.ibm.com/support/docview.wss?rs=171&uid=swg21433474 ).
| Quote: |
| I'd assumed that by now it would be ok, but from what you say it still sounds a combination of FixPacks, interim fixes and crossed fingers.... |
I don't understand. The link above shows "Testing failed, see note 1" in the cells relating to Solaris, so there's no ambiguity - it doesn't pass the testing, so is not supported.
| Quote: |
| Multi-Instance seems to offer a 'faster failover' HA solution (compared to the VCS approach), but it doesn't sound like it's 'production ready' on Sun Solaris? |
Fair comment, though from that web page the defect does not appear to be in IBM code. |
|
| Back to top |
|
|
| CHAFF |
| Posted: Thu Mar 31, 2011 2:03 pm Post subject: Re: Multi instance on solaris |
|
|
Newbie
Joined: 07 Mar 2011
Posts: 5
|
| indian82 wrote: |
Hi everubody,
Thanks evrybody for the things you put on this thread...really got lot of information.But just want to know has the problem with NSF V4 has been debugged and did anybody do mq v7 multi instance + solaris + contact admin/NSF V4 and were succesfull and went to production.Thank you guys. |
Because the issue seems to lie on the Solaris client side NFSv4, there is no ETA as to when the defect will be fixed. Because MQMI+contact admin+NSFv4 is such a broad topic, it is very difficult to find exact information regarding this setup. Its basically "put it in and pray" that you're not affected by the Solaris+NSFv4 bug. |
|
| Back to top |
|
|
| mvic |
| Posted: Thu Mar 31, 2011 2:47 pm Post subject: Re: Multi instance on solaris |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| CHAFF wrote: |
| Its basically "put it in and pray" that you're not affected by the Solaris+NSFv4 bug. |
Why? Why put it in at all, if there's a bug? |
|
| Back to top |
|
|
| CHAFF |
| Posted: Thu Mar 31, 2011 2:51 pm Post subject: Re: Multi instance on solaris |
|
|
Newbie
Joined: 07 Mar 2011
Posts: 5
|
| mvic wrote: |
| CHAFF wrote: |
| Its basically "put it in and pray" that you're not affected by the Solaris+NSFv4 bug. |
Why? Why put it in at all, if there's a bug? |
new feature/enhancement = testing testing testing |
|
| Back to top |
|
|
| paustin_ours |
| Posted: Thu Jul 14, 2011 7:17 am Post subject: |
|
|
Yatiri
Joined: 19 May 2004
Posts: 667
Location: columbus,oh
|
This topic i see has digressed from the earlier discussion about whether or not MC91 is the way to go for MQ HA. It looked liked for multiinstance, the MQ would handle the healthcheck and what not but i dont see any clear documentation on how to get it to work for traditional HA.
thoughts? |
|
| Back to top |
|
|
|
|
