| Author |
Message
|
| exerk |
 Posted: Tue Nov 24, 2009 1:49 pm Post subject: |
 |
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| JosephGramig wrote: |
| exerk wrote: |
| Methinks that there should be a clear delineation between HA and MI queue managers |
HA swings resources over like IP address and other file systems.
MI locks access to a shared file system and uses a different IP.
You can still do both despite what you see for MC91. |
Agreed. But what is true HA? I've just had a 2 server HA solution reduced to one thanks to some numbnuts blowing the OS on one of them, so now we're looking to make it a 3 server solution, something we can't do with a MI queue manager. Hence my remark about clear delineation - it can't be too far down the road before people start referring to MI queue managers as an HA solution when clearly it's not (pseudo HA in my humble opinion).
_________________
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 |
|
 |
| bruce2359 |
| Posted: Tue Nov 24, 2009 1:56 pm Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9482
Location: US: west coast, almost. Otherwise, enroute.
|
If all this HA and MI stuff gets too complicated, it's time for a WMQ on a z/OS mainframe, where Parallel Sysplex has already been invented.
_________________
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 |
|
|
| Michael Dag |
| Posted: Tue Nov 24, 2009 2:38 pm Post subject: |
|
|
Jedi Knight
Joined: 13 Jun 2002
Posts: 2607
Location: The Netherlands (Amsterdam)
|
HA comes in many shapes ... depending on the number of 9's in uptime you require ... The only 'fully uninterupted service' from an MQ perspective would be shared queueing on z/OS as Bruce pointed out.
_________________
Michael
MQSystems Facebook page |
|
| Back to top |
|
|
| George Carey |
| Posted: Wed Nov 25, 2009 9:59 am Post subject: HA or MI |
|
|
Knight
Joined: 29 Jan 2007
Posts: 500
Location: DC
|
Yes, MI ... shared volumes but no IP moves.
But with new client functionality VIP become less and less necessary.
But in any case I am talking about MI. MI configured but in the case of this problem no MI running.
All QMGRs down but one and only one instance running on the one server via the command: strmqm QMGRname
GTC
_________________
"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: Wed Nov 25, 2009 10:09 am Post subject: HA or no HA |
|
|
Knight
Joined: 29 Jan 2007
Posts: 500
Location: DC
|
| Quote: |
| ....so now we're looking to make it a 3 server solution, something we can't do with a MI queue manager. ... |
Why can you not have 3 or more ... If I understand the strmqm -x QMGR command ... you can have as many as you like standby QMGRs and when the primary is determined to have failed the first standby QMGR to grab exclusive use of the shared volumes will take over and become active.
I do not recall seeing/reading you can have only one(1) standby QMGR.
_________________
"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: Wed Nov 25, 2009 10:14 am Post subject: z/os simplifying |
|
|
Knight
Joined: 29 Jan 2007
Posts: 500
Location: DC
|
| Quote: |
| ... If all this HA and MI stuff gets too complicated, it's time for a WMQ on a z/OS mainframe, ... |
Yeah, whenever I think I need to simplify and make my I/T life less complicated, I always think about going to Z/OS !!!???!!!
_________________
"Truth is ... grasping the virtually unconditioned",
Bernard F. Lonergan S.J.
(from book titled "Insight" subtitled "A Study of Human Understanding") |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Wed Nov 25, 2009 10:18 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9482
Location: US: west coast, almost. Otherwise, enroute.
|
Whoop, whoop, whoop! sarcasm alarm!
_________________
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 |
|
|
| Vitor |
| Posted: Wed Nov 25, 2009 10:24 am Post subject: Re: z/os simplifying |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| George Carey wrote: |
| Quote: |
| ... If all this HA and MI stuff gets too complicated, it's time for a WMQ on a z/OS mainframe, ... |
Yeah, whenever I think I need to simplify and make my I/T life less complicated, I always think about going to Z/OS !!!???!!! |
I've always found z/OS to be the ideal OS - simple, uncomplicated and utterly reliable. As well as having loads of processing power.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| exerk |
| Posted: Wed Nov 25, 2009 11:31 am Post subject: Re: HA or no HA |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| George Carey wrote: |
..Why can you not have 3 or more ... If I understand the strmqm -x QMGR command ... you can have as many as you like standby QMGRs and when the primary is determined to have failed the first standby QMGR to grab exclusive use of the shared volumes will take over and become active.
I do not recall seeing/reading you can have only one(1) standby QMGR. |
Then maybe I am misinterpreting the Info Centre, which states (in multiple places "...a pair of servers..." or "...both servers...", and what nails it for me "...Note: Only two queue manager instances can run at the same time...", which implies that if both boxes disappear, and you can have it on other servers, that it is not going to pick up automatically as 'true' HA would.
Time to create another VM and try it I think 
_________________
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 |
|
|
| bruce2359 |
| Posted: Wed Nov 25, 2009 11:43 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9482
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
| I do not recall seeing/reading you can have only one(1) standby QMGR. |
I recall this discussed in the webcast roll-out of MI last week.
The way it was worded was something like you can create multiple standbys, but MI only supports one active and one standby - at any given moment. If the active fails, the standby becomes the active; and you (or automation) can alter one of the standby standbys to become the standby.
The recorded webcast should be available for replay by now.
_________________
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 |
|
|
| George Carey |
| Posted: Wed Nov 25, 2009 12:57 pm Post subject: MI trys |
|
|
Knight
Joined: 29 Jan 2007
Posts: 500
Location: DC
|
| Quote: |
| "... Time to create another VM and try it I think ..." |
Same here, to confirm one way or other.
_________________
"Truth is ... grasping the virtually unconditioned",
Bernard F. Lonergan S.J.
(from book titled "Insight" subtitled "A Study of Human Understanding") |
|
| Back to top |
|
|
| mvic |
| Posted: Wed Nov 25, 2009 1:58 pm Post subject: |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| bruce2359 wrote: |
| you can create multiple standbys, but MI only supports one active and one standby - at any given moment. If the active fails, the standby becomes the active; and you (or automation) can alter one of the standby standbys to become the standby. |
This is correct. A multi-instance queue manager is only one queue manager, but with the capability for a "standby" process to be started on a 2nd machine awaiting the chance to run. When it gets the chance to run (because of failure/switchover), it becomes *the* queue manager.
To simulate the traditional HA managed situation, one can then - as soon as data about the first failure is collected from the original machine and the green light is given - make the original machine run a "standby" process, awaiting failure/switchover of what is now *the* queue manager. |
|
| Back to top |
|
|
| exerk |
| Posted: Wed Nov 25, 2009 3:08 pm Post subject: Re: MI trys |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| George Carey wrote: |
| Quote: |
| "... Time to create another VM and try it I think ..." |
Same here, to confirm one way or other. |
Indeed, and find an easy way to very quickly remap the DNS name of the failed machine to that of the standby standby - only two IP Addresses/DNS names remember. I still think 'traditional' HA has the edge here.
| mvic wrote: |
| To simulate the traditional HA managed situation, one can then - as soon as data about the first failure is collected from the original machine and the green light is given - make the original machine run a "standby" process, awaiting failure/switchover of what is now *the* queue manager. |
Assuming you get that box back. As I stated in a previous post, somebody blew the OS on one of the servers in an HA setup. Easier to set up another box to accept the resource group than jump through the hoops of shared storage, DNS remaps etc.
As far as i am concerned, MI queue managers have yet to mature (let's face it, the paint is still wet) and I would only use them at the moment for low-level solutions where an enhanced availability is required.
_________________
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 |
|
|
| mvic |
| Posted: Wed Nov 25, 2009 3:22 pm Post subject: Re: MI trys |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| exerk wrote: |
| mvic wrote: |
| To simulate the traditional HA managed situation, one can then - as soon as data about the first failure is collected from the original machine and the green light is given - make the original machine run a "standby" process, awaiting failure/switchover of what is now *the* queue manager. |
Assuming you get that box back. As I stated in a previous post, somebody blew the OS on one of the servers in an HA setup. Easier to set up another box to accept the resource group than jump through the hoops of shared storage, DNS remaps etc. |
OK. Similar should be do-able relatively quickly with the new 7.0.1 dspmqinf and addmqinf, and once the NFS exporting/mounting has been done.
| Quote: |
| As far as i am concerned, MI queue managers have yet to mature (let's face it, the paint is still wet) and I would only use them at the moment for low-level solutions where an enhanced availability is required. |
Fair comment, indeed. |
|
| Back to top |
|
|
| George Carey |
| Posted: Wed Nov 25, 2009 3:30 pm Post subject: Use MQ client with MI |
|
|
Knight
Joined: 29 Jan 2007
Posts: 500
Location: DC
|
If you stick to MQclients connectivity to your MQ server, the MI setup could be ideal.
You have under MQv7.0.1 syntax like this for clients channel setup.
def chl(chan1) chltype(clntconn) trptype(tcp) +
conname('server1(123),server2(345)') qmname(QMGRxyz)
A v7.0.1 client will auto reconnect using the conname servers and ports in a round robin fashion!!! (no need for VIP)
The servers 1 and 2 will be using the same Q data on reconnection by having /var/mqm as a shared contact admin volume that is NFS mounted on both servers, one running
QMGRxyz as primary and one is standby and your done!!!
Where is the hassle with that ??
_________________
"Truth is ... grasping the virtually unconditioned",
Bernard F. Lonergan S.J.
(from book titled "Insight" subtitled "A Study of Human Understanding")
Last edited by George Carey on Mon Nov 30, 2009 11:03 am; edited 1 time in total |
|
| Back to top |
|
|
|
|
