| Author |
Message
|
| smdavies99 |
 Posted: Thu Feb 20, 2014 11:31 pm Post subject: IIB/WMB in a Hyper-V environment |
 |
|
Jedi Council
Joined: 10 Feb 2003
Posts: 6076
Location: Somewhere over the Rainbow this side of Never-never land.
|
Has anyone been looking at or done this? IF so what are your feelings/experiences etc?
_________________
WMQ User since 1999
MQSI/WBI/WMB/'Thingy' User since 2002
Linux user since 1995
Every time you reinvent the wheel the more square it gets (anon). If in doubt think and investigate before you ask silly questions. |
|
| Back to top |
|
 |
| JosephGramig |
| Posted: Fri Feb 21, 2014 8:12 am Post subject: |
|
|
Grand Master
Joined: 09 Feb 2006
Posts: 1244
Location: Gold Coast of Florida, USA
|
So you have a requirement where you would need to quickly generate additional brokers to handle load?
Like a national "Healthcare Web Site" application?
Where the load is unknown before it is needed. |
|
| Back to top |
|
|
| DirtyHowi |
| Posted: Fri Feb 21, 2014 2:43 pm Post subject: Re: IIB/WMB in a Hyper-V environment |
|
|
Novice
Joined: 17 Feb 2014
Posts: 22
|
| smdavies99 wrote: |
| Has anyone been looking at or done this? IF so what are your feelings/experiences etc? |
all of our servers are hyper-v, runs just like its on the desktop, if the servers are provisioned correctly. My workstation is hyper-v and i have no problems with it.
now load balancing is another issue altogether |
|
| Back to top |
|
|
| smdavies99 |
| Posted: Fri Feb 21, 2014 6:16 pm Post subject: |
|
|
Jedi Council
Joined: 10 Feb 2003
Posts: 6076
Location: Somewhere over the Rainbow this side of Never-never land.
|
The main issue I see is the load balancing.
How are you stopping the 'messages overtaking' syndrome?
By this I mean,
A message is sent with an alteration to something
A second message is send almost immediately reversing that something.
The first message is is processed by a broker in a VM that suddenly experiences heavy load.
The second message is processed in another VM and get to the endpoint BEFORE the first message.
I call this indirect message affinity.
The two messages are not linked in that there is no sequence number in them but the data they contain is linked and if they are received at the ultimate destination out of order then all sorts of chaos could enschew,
This is just like any Active-Active scenario but the MS guys seems to think that Hyper-V is the be-all and end-all of system provisioning.
They baulk at us old fogeys demanding the good old days.
(this is a different customer to my MSCS clustering issue btw). They have Hyper-V there but can't provision it correctly so they have wisely decided to go with physical servers for production.
_________________
WMQ User since 1999
MQSI/WBI/WMB/'Thingy' User since 2002
Linux user since 1995
Every time you reinvent the wheel the more square it gets (anon). If in doubt think and investigate before you ask silly questions. |
|
| Back to top |
|
|
| Gralgrathor |
| Posted: Sat Feb 22, 2014 3:09 am Post subject: |
|
|
Master
Joined: 23 Jul 2009
Posts: 297
|
| smdavies99 wrote: |
| I call this indirect message affinity. |
How would you deal with this?
_________________
A measure of wheat for a penny, and three measures of barley for a penny; and see thou hurt not the oil and the wine. |
|
| Back to top |
|
|
| smdavies99 |
| Posted: Sat Feb 22, 2014 4:18 am Post subject: |
|
|
Jedi Council
Joined: 10 Feb 2003
Posts: 6076
Location: Somewhere over the Rainbow this side of Never-never land.
|
I'd go for a an Active Passive solution.
What about you?
How would you deal with it?
_________________
WMQ User since 1999
MQSI/WBI/WMB/'Thingy' User since 2002
Linux user since 1995
Every time you reinvent the wheel the more square it gets (anon). If in doubt think and investigate before you ask silly questions. |
|
| Back to top |
|
|
| Gralgrathor |
| Posted: Sat Feb 22, 2014 4:57 am Post subject: |
|
|
Master
Joined: 23 Jul 2009
Posts: 297
|
| smdavies99 wrote: |
| I'd go for a an Active Passive solution. |
Will that guarantee ordered delivery and processing? I'm not sure how that would work.
| smdavies99 wrote: |
| How would you deal with it? |
I'm not really at home in clustered MQ environments and configuration. Still a lot of learning to do. My first impulse would be to see if there's some programmatic fix to the problem. But it would be better if it could be fixed through a proper configuration of MQ.
_________________
A measure of wheat for a penny, and three measures of barley for a penny; and see thou hurt not the oil and the wine. |
|
| Back to top |
|
|
| Michael Dag |
| Posted: Sat Feb 22, 2014 5:07 am Post subject: |
|
|
Jedi Knight
Joined: 13 Jun 2002
Posts: 2607
Location: The Netherlands (Amsterdam)
|
you cannot avoid the 'message overtaking' or message affinity, how well you design you envirnoment it WILL bite you at your worst moment...
Applications should handle 'order' issues by exception handling or holding data for reprocessing...
_________________
Michael
MQSystems Facebook page |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Sat Feb 22, 2014 9:05 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
And this isn't really related to the title of this thread, "IIB/WMB in a Hyper-V environment".
You have the some consideration in any environment where multiple brokers are running in parallel, regardless of how those individual Brokers are implemented - single stand alone physical servers, single stand alone virtual servers, Active/Passive set up using Multi Instance or 'classic' H.A. solutions like MCSC or VCS. If you have 2 or more Brokers getting traffic load balanced to them by MQ Clustering for MQ messages and IP load balancers for HTTP, you face problems when apps want message affinity.
Lately we are struggling with this - people read about new nodes like DatabaseInput, Sequence and Resquence and these things just don't work out of the box in a multi broker environment. They look great when you only have a single broker, but it gets complicated fast when you have multiple brokers and apps insist on message order.
FileInput Node - we got one flow that starts like this. They only have it active on one of our Brokers. The flow is stopped on the other Broker. If the first Broker's H.A. support fails it and that Broker is down for the count, the app team manually starts up the flow in the second Broker. Manual H.A., ugh. But what are you going to do?
If they can't eliminate the message affinity, you have to either force them to use only one of your multiple Brokers and now they aren't take full advantage of the high availability of your multi broker design, or you have to add code into the flows so they all refer to some single entity (a H.A. remote database) that can coordinate all this.
I know lancelotlinc was a big proponent of multiple active brokers and to heck with that Activ/Passive stuff that causes more outages then it prevents, I wonder how they deal with this. Maybe in their shangri-la environment all the kids are above average and message affinity does not exist. lancelotlinc, where's he been?
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Sat Feb 22, 2014 5:45 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
LancelotLinc has probably implemented a singleton in the cache, retrieving the stub for the RMI to then internally sequence the messages through the singleton.... 
Kidding apart the best design is still to dump it into a db and process on a single broker when the sequence is ready... 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| smdavies99 |
| Posted: Sun Feb 23, 2014 6:41 am Post subject: |
|
|
Jedi Council
Joined: 10 Feb 2003
Posts: 6076
Location: Somewhere over the Rainbow this side of Never-never land.
|
There is an issue where a customer picks up on some technology, in this case, Hyper-V and adopts it lock stock and steaming cess-pit.
That is the instance I am fighting. He wants to spit his broker capacity between two active nodes rather than the Active-Pasive setup we are recommending.
If in the Active-Active situation a message does overtake another then the consequences will be severe but not life threatening. A lot of people will be very unhappy and a good few companies will lose a whole shed load of cash but the people making this decision are Civil Servants who have no business aptitude whatsoever. We will probably have to go with Active-Active and with no control over the behaviour of the Load Balance.
Not ideal but the size of the contract means that we can't just walk away.
Thanks to the inputs on this thread.
_________________
WMQ User since 1999
MQSI/WBI/WMB/'Thingy' User since 2002
Linux user since 1995
Every time you reinvent the wheel the more square it gets (anon). If in doubt think and investigate before you ask silly questions. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Sun Feb 23, 2014 3:45 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| smdavies99 wrote: |
There is an issue where a customer picks up on some technology, in this case, Hyper-V and adopts it lock stock and steaming cess-pit.
That is the instance I am fighting. He wants to spit his broker capacity between two active nodes rather than the Active-Pasive setup we are recommending.
If in the Active-Active situation a message does overtake another then the consequences will be severe but not life threatening. A lot of people will be very unhappy and a good few companies will lose a whole shed load of cash but the people making this decision are Civil Servants who have no business aptitude whatsoever. We will probably have to go with Active-Active and with no control over the behaviour of the Load Balance.
Not ideal but the size of the contract means that we can't just walk away.
Thanks to the inputs on this thread. |
If you are using 7.5 did you think about using defbind(group) instead of (notfixed)?
_________________
MQ & Broker admin |
|
| Back to top |
|
|
|
|
