| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| How cluster configuration information will be exchanged |
« View previous topic :: View next topic » |
| Author |
Message
|
| raj429 |
 Posted: Wed Jan 12, 2011 1:20 am Post subject: How cluster configuration information will be exchanged |
 |
|
Acolyte
Joined: 04 Jul 2009
Posts: 53
|
I have created a setup of 4 queue managers..
QM1 & QM2 FRs and 2 PRs QM3 & QM4
cluster sender channel definitions as below..
QM3--> QM1
QM4--> QM2
now on QM3, I can see CLUSSDR for bothe QM1 & QM2
on QM4, CLUSSDR for QM1 & QM2.
now If i created a Cluster ql(XX) in QM3, will that be updated to QM2 by Auto defined channel of QM3-->QM2 or QM3-->QM1-->QM2?
or QM2 gets the updates from both?
I am trying to understand the flow of configuration information exchange..
any suggestions |
|
| Back to top |
|
 |
| mqjeff |
| Posted: Wed Jan 12, 2011 1:35 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
If an application is connected to a queue manager that is part of a cluster issues an MQOPEN for a queue that is A) shared in the cluster AND B) not local AND C) not already known to the queue manager, THEN the queue manager requests the information about the queue from a full repository.
That is how the information flows.
Once that queue name has been RESOLVED and that the final destination is CHOSEN, then the messages that are MQPUT go DIRECTLY to the final destination queue manager over the autodefined channel.
Review the Clusters manual. |
|
| Back to top |
|
|
| raj429 |
| Posted: Wed Jan 12, 2011 2:12 am Post subject: |
|
|
Acolyte
Joined: 04 Jul 2009
Posts: 53
|
update to my question below, is there any possibility of duplicate MQ configuration information in SCRQ of a FR?
in cluster, I can see the SCRQ curdepth is different for bother FRs. |
|
| Back to top |
|
|
| mqjeff |
| Posted: Wed Jan 12, 2011 2:13 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
There is not a one to one correlation between messages on SCRQ and the number of objects in the cluster.
Do not assume anything about the contents of messages in SCRQ. This is an internal queue and its format and use is not published. |
|
| Back to top |
|
|
| JosephGramig |
| Posted: Wed Jan 12, 2011 1:18 pm Post subject: |
|
|
Grand Master
Joined: 09 Feb 2006
Posts: 1244
Location: Gold Coast of Florida, USA
|
Also, every FR must have an explicit cluster sender channel to every other FR in the same cluster for the FRs in the cluster to get updates.
So
QM1->QM2 explicit cluster sender
QM2->QM1 explicit cluster sender
If you were not this way before, then demote QM2. Then refresh cluster(...) repos(yes). Then make sure you have the channels as described above. Then promote QM2 to be an FR. Then refresh cluster(...) repos(yes) at QM3 and QM4, just for good measure.
FRs will only forward updates to the cluster to other FRs to which it has an explicit cluster sender to them. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Wed Jan 12, 2011 1:39 pm Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9470
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
| is there any possibility of duplicate MQ configuration information in SCRQ of a FR? |
I suppose it depends on your definition of duplicate.
Yes, there will be information about cluster object definitions that exist on queue managers. So, if you have 4 queues with the identical queue name in your cluster, you will have definitions for them.
Another case where there will be duplicate cluster into: If you change an attribute of one of the cluster objects, due to latency in FR updates, you will briefly have definitions in different attributes in FRs, and likely PRs. This is normal behavior.
[edit]
Briefly, very briefly.
When a cluster object is defined, it is assigned a sequence number (like 0001), and the definition is included in the local repository (where the object was defined). It is then sent to the FR, which will send it to its peer FR. Each repos gets the same sequence number instance of the object.
When this same object is in any way altered (or deleted, or taken out of the cluster), the sequence number is incremented by 1; and the cluster object information is sent again to the FR, which then sends it to the partner FR. If any PR has expressed interest, it will be sent to the PR(s), as well.
[/edit]
_________________
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 |
|
|
| raj429 |
| Posted: Mon Jan 17, 2011 1:09 am Post subject: |
|
|
Acolyte
Joined: 04 Jul 2009
Posts: 53
|
Thanks Bruce, got the idea to some extent..
| Quote: |
When a cluster object is defined, it is assigned a sequence number (like 0001), and the definition is included in the local repository (where the object was defined). It is then sent to the FR, which will send it to its peer FR. Each repos gets the same sequence number instance of the object.
|
In FR, Cluster information will be exchanged only by manual defined CLUSSDR channels right?
In my setup, I brought another full repository QM5 connecting to QM2(also a FR).
1. Do I have to connect(manual) QM5 FR to another QM1 FR also i.e, to both QM2 & QM1?
2. IN PRs, I had seen auto defined cls channels created for this QM5 instantly and running mode. where as that PR, is directly(manually ) connected to QM2. why these auto channels to QM5? any configuration info exchange from PRs?(i am testing this, trying again..it got messed up earliar  appreciated your views on this).
2.a. If yes(config exchange from PR directly to all FRs), then y again conneecting all FR/ |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Mon Jan 17, 2011 7:00 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9470
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
| In FR, Cluster information will be exchanged only by manual defined CLUSSDR channels right? |
To function as an FR, there needs to be a manually defined clussdr to another FR. So, yes.
| Quote: |
| In my setup, I brought another full repository QM5 connecting to QM2(also a FR). |
| Quote: |
| 1. Do I have to connect(manual) QM5 FR to another QM1 FR also i.e, to both QM2 & QM1? |
No.
| Quote: |
| 2. IN PRs, I had seen auto defined cls channels created for this QM5 instantly and running mode. where as that PR, is directly(manually ) connected to QM2. why these auto channels to QM5? |
| Quote: |
...any configuration info exchange from PRs[?/quote]
Yes, PR qmgrs send info about its clustered objects to the FR. |
Happy to share.
| Quote: |
| (i am testing this, trying again..it got messed up earliar appreciated your views on this). |
Happy to share.
| Quote: |
| 2.a. If yes (config exchange from PR directly to all FRs), |
PRs send cluster info to its FR. The FR will then send cluster info to its partner FR.
| Quote: |
| then y again conneecting all FR |
Are you asking why aren't all repositories (PRs and FRs) all permanently interconnected? If so, one of the major benefits of clusters is that connections are only required when and if applications need them, and the channels are torn down when no longer needed.
_________________
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 |
|
|
| fjb_saper |
| Posted: Mon Jan 17, 2011 11:20 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
| raj429 wrote: |
Thanks Bruce, got the idea to some extent..
In FR, Cluster information will be exchanged only by manual defined CLUSSDR channels right?
In my setup, I brought another full repository QM5 connecting to QM2(also a FR).
1. Do I have to connect(manual) QM5 FR to another QM1 FR also i.e, to both QM2 & QM1?
|
Yes. AFAIK you need to define a manual cluster sender channel from each FR to each other FR. So if you have 3 FR's in your cluster each FR needs 2 manual cluster senders to be defined...
Be reminded however that best practice tells us to define only 2 FRs
Usually and additional FR only gets defined for reasons of geographical independence and such... 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Mon Jan 17, 2011 11:58 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9470
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
1. Do I have to connect(manual) QM5 FR to another QM1 FR also i.e, to both QM2 & QM1?
No. |
I was not clear in my answer. Sorry about that.
I was answering the question as to whether an FR needs to have explicit clussdrs defined both (the other two) FRs. So, no. Not necessary, not required.
As with all qmgrs in the cluster, an FR only needs an explicit clussdr to its (peer) FR. In this sense, yes - an FR must have an explicit clussdr to a(nother) FR.
_________________
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 |
|
|
| raj429 |
| Posted: Thu Jan 20, 2011 12:46 am Post subject: |
|
|
Acolyte
Joined: 04 Jul 2009
Posts: 53
|
| Quote: |
raj429 wrote:
Thanks Bruce, got the idea to some extent..
In FR, Cluster information will be exchanged only by manual defined CLUSSDR channels right?
In my setup, I brought another full repository QM5 connecting to QM2(also a FR).
1. Do I have to connect(manual) QM5 FR to another QM1 FR also i.e, to both QM2 & QM1?
fjp_Saper wrote:
Yes. AFAIK you need to define a manual cluster sender channel from each FR to each other FR. So if you have 3 FR's in your cluster each FR needs 2 manual cluster senders to be defined...
|
| Quote: |
Bruce2359 wrote:
I was answering the question as to whether an FR needs to have explicit clussdrs defined both (the other two) FRs. So, no. Not necessary, not required.
As with all qmgrs in the cluster, an FR only needs an explicit clussdr to its (peer) FR. In this sense, yes - an FR must have an explicit clussdr to a(nother) FR.
|
Saper & bruce thanks for inputs..but I guess answeres are contradicting
I belive it should be ideally, QM1 & QM2 & QM5 are full repositories..then
QM1 should have 2 Manual clussdrs to QM2 & QM5 ;
QM2 should have 2 manual clussdrs to QM1 & QM5 ;
QM5 should have 2 manual clussdrs to QM1 & QM2 ;
Coz, if any Qmgr went down, other 2 FRs can communicate..
other scenario..
if QM1 is Manually connected only to QM2
QM2 is Manually connected to only QM5
QM5 is Manuallyconnected to only QM2
Now if QM2 is down, will QM1 & QM5 able to excahnge information? I guess nope...
Please correct me if my understanding is wrong |
|
| Back to top |
|
|
| mqjeff |
| Posted: Thu Jan 20, 2011 4:18 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
They will be able to communicate using autodefined channels.
You should not have 3 FRs. |
|
| Back to top |
|
|
| Vitor |
| Posted: Thu Jan 20, 2011 5:18 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| raj429 wrote: |
| Please correct me if my understanding is wrong |
It's wrong. Re-read these posts & the clustering manual.
Do not have more than 2 FRs without a really, really, exceptionally good reason.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
