Author |
Message
|
Zappa |
Posted: Wed Nov 10, 2010 5:08 am Post subject: Multi-Instance & SAN not NFS |
|
|
 Acolyte
Joined: 06 Oct 2005 Posts: 55 Location: UK
|
When creating multi-instance queue managers can the log & data shared storage be mounted from a SAN concurrently to both servers or does it have to be via an NFS server? |
|
Back to top |
|
 |
JLRowe |
Posted: Wed Nov 10, 2010 5:21 am Post subject: |
|
|
 Yatiri
Joined: 25 May 2002 Posts: 664 Location: South East London
|
Generally speaking, a SAN will not let you mount the same volume on more than one server, a SAN is working at a lower layer than the filesystem above it.
In the older style of clustering using active-inactive, a SAN volume would be failed over between two or more boxes using a product such as sun cluster or MSCS.
The newer shared queue managers operate at the higher file system level, and therefore two queue managers know how to interact & failover with each other using the filesystem as the means of communication. This is a lot easier to configure and get working as you are operating at the filesystem level.
Sorry, could explain it a bit more clearly, but hope you get the idea. |
|
Back to top |
|
 |
Zappa |
Posted: Wed Nov 10, 2010 5:54 am Post subject: |
|
|
 Acolyte
Joined: 06 Oct 2005 Posts: 55 Location: UK
|
Thanks for that, which is fairly much what I thought too but had to ask.
So the NFS server becomes a single point of failure if it is not HA'd and if it was HA'd then failing over the NFS share would cause an outage to the queue manager that is meant to be highly available. |
|
Back to top |
|
 |
JLRowe |
Posted: Wed Nov 10, 2010 6:00 am Post subject: |
|
|
 Yatiri
Joined: 25 May 2002 Posts: 664 Location: South East London
|
Yes, but the value premise is that it is a whole lot easier to make a filesystem highly avalable. I believe that the queue manager will keep retrying access to the file system, meaning it will bounce back up again when the file system is available. |
|
Back to top |
|
 |
bruce2359 |
Posted: Wed Nov 10, 2010 7:47 am Post subject: |
|
|
 Poobah
Joined: 05 Jan 2008 Posts: 9469 Location: US: west coast, almost. Otherwise, enroute.
|
Quote: |
So the NFS server becomes a single point of failure if it is not HA'd |
Oddly worded.
NFS v4 is required for MI.
As I interpret the MI deployment (subject to misunderstanding) NFS software executes on both qmgrs in the MI. Each NFS instance monitors the filesystem lock - which the active qmgr owns until the active qmgr fails. _________________ 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 |
|
 |
JLRowe |
Posted: Thu Nov 11, 2010 8:45 am Post subject: |
|
|
 Yatiri
Joined: 25 May 2002 Posts: 664 Location: South East London
|
Yes, but you still need a separate NFS server - which becomes the point of failure? |
|
Back to top |
|
 |
mqjeff |
Posted: Thu Nov 11, 2010 9:05 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008 Posts: 17447
|
JLRowe wrote: |
Yes, but you still need a separate NFS server - which becomes the point of failure? |
You need a SAN device or server of some kind as well, yes? |
|
Back to top |
|
 |
Zappa |
Posted: Fri Nov 12, 2010 3:34 am Post subject: |
|
|
 Acolyte
Joined: 06 Oct 2005 Posts: 55 Location: UK
|
The MQ systems mounting the file system be it NFS or SMB are both acting as the client to the NFS or SMB server. The shared filesystem could be sourced from a SAN which should have its own fault tolerances and redundancies and shouldn't be the SPOF.
I'm thinking that whatever serves this file share is the SPOF and will need to be HA’d if you want the queue manager to be HA.
In my mind (please correct me if I’m mislead) MI just moves the need for a HA cluster from the MQ servers to the file server/s. |
|
Back to top |
|
 |
JLRowe |
Posted: Fri Nov 12, 2010 7:26 am Post subject: |
|
|
 Yatiri
Joined: 25 May 2002 Posts: 664 Location: South East London
|
Yes, that's correct, it just 'pushes' the HA requirement down from the MQ app level to the file system level. But, the reason to do this is it's a lot easier to make a file system highly available. |
|
Back to top |
|
 |
mqjeff |
Posted: Fri Nov 12, 2010 8:21 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008 Posts: 17447
|
It's a lot easier to make a file system highly available and it's a lot more common to already have any NFS or SAN service already be highly available before MQ enters the picture at all. |
|
Back to top |
|
 |
fjb_saper |
Posted: Fri Nov 12, 2010 12:42 pm Post subject: |
|
|
 Grand High Poobah
Joined: 18 Nov 2003 Posts: 20756 Location: LI,NY
|
mqjeff wrote: |
It's a lot easier to make a file system highly available and it's a lot more common to already have any NFS or SAN service already be highly available before MQ enters the picture at all. |
I think you meant contact admin service. As I remember it contact admin can attach the same volume to multiple boxes whereas SAN will limit the volume to one box...
But then what do I know I have to ask those network gurus every time...  _________________ MQ & Broker admin |
|
Back to top |
|
 |
mqjeff |
Posted: Fri Nov 12, 2010 1:08 pm Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008 Posts: 17447
|
NFS is something different.
Regardless of whether or not a Storage Area Network volume is mounted on a single computer at a time or not, the Storage Area Network controller/server can still be Highly Available (and usually is). |
|
Back to top |
|
 |
|