Is it advisable to have a load balancer name in the Connection Name?

Why am I asking this?

if you have 2 QMGRs on the same host using the same port, how would the load balancer know which QMGR to start the channel on?

What if you have 2 QMGRs on different host with the same port behind the same load balancer?

If so, must one populate the local communication Address as well?

Is it advisable to have a load balancer name in the Connection Name?

Why am I asking this?

if you have 2 QMGRs on the same host using the same port, how would the load balancer know which QMGR to start the channel on?

What if you have 2 QMGRs on different host with the same port behind the same load balancer?

If so, must one populate the local communication Address as well?

Hi Emile.

No, it's not recommended to attempt to load balance QM to QM connections of any kind in this way. 

For a client (application) to QMgr connection, load balancing can be viable if correctly configured.  However, QM to QM channel protocols require that connections always reach the exact same QM (as you describe).  This is crucial to confirming message/batch integrity, ordering etc.  therefore you will receive errors if the connection is incorrectly routed.
The only case in which some form of 'redirection' (I won't say 'balancer') would make sense is if this is an HA environment and the two 'Queue Managers' are in fact instances of the exact same queue manager (Multi Instance, RDQM, OS level clustering etc.)
If you can describe what you are trying to achieve with such a configuration I or others here may be able to suggest better alternative approaches?

Regards

Anthony

Is it advisable to have a load balancer name in the Connection Name?

Why am I asking this?

if you have 2 QMGRs on the same host using the same port, how would the load balancer know which QMGR to start the channel on?

What if you have 2 QMGRs on different host with the same port behind the same load balancer?

If so, must one populate the local communication Address as well?

Hi Anthony,

Thanks for your response.

I do not like the setup but this is how I found it. 

I see " AMQ9407W: Cluster queue <Queue Name> is defined inconsistently"  AND "AMQ9456I: Update not received for object
<Queue Name>, queue manager <QMGR NAME> from full repository for cluster <CLUSTER NAME> " very often in my logs.

Here is another " AMQ9469W: Update not received for CLUSRCVR channel TO.QMGR NAME hosted on queue
manager QMGR NAME in cluster CLUSTER NAME

.

When I looked, I see at CLUSSDR CHL, it has the LOAD BALANCER in the CONNAME, hence my question.

Hope I cleared some misunderstandings and you can advise

Thanks Emile

Hi Emile.

No, it's not recommended to attempt to load balance QM to QM connections of any kind in this way. 

For a client (application) to QMgr connection, load balancing can be viable if correctly configured.  However, QM to QM channel protocols require that connections always reach the exact same QM (as you describe).  This is crucial to confirming message/batch integrity, ordering etc.  therefore you will receive errors if the connection is incorrectly routed.
The only case in which some form of 'redirection' (I won't say 'balancer') would make sense is if this is an HA environment and the two 'Queue Managers' are in fact instances of the exact same queue manager (Multi Instance, RDQM, OS level clustering etc.)
If you can describe what you are trying to achieve with such a configuration I or others here may be able to suggest better alternative approaches?

Regards

Anthony

Is it advisable to have a load balancer name in the Connection Name?

Why am I asking this?

if you have 2 QMGRs on the same host using the same port, how would the load balancer know which QMGR to start the channel on?

What if you have 2 QMGRs on different host with the same port behind the same load balancer?

If so, must one populate the local communication Address as well?

Hi Emile

To advise on how this setup should be modified would really need an understanding of what it was originally trying to achieve.  Presumably somebody believe rightly or wrongly that this was giving an element of load balancing or high availability to the configuration.  So all I can really do is advise in general terms...

A Cluster Receiver channel should usually only be configured with the hostname/IP of the exact host for that QM (or a comma separated list if using a failover technology).  This will be used to autodefine sender channels as I think you  already understand.  A CLUSSDR channel should only ever point to (one) full repository.

If load balancing of messages is required, this can be configured at the cluster/destination level using the WLM properties on queues, channels etc... the QM to QM connections themselves should not be load balanced.  (More recently there is also the built in option of load balancing client application connections - Uniform Clusters - but it doesn't sound like that was what was being attempted here.)

Hope this helps a little,

Regards

Hi Anthony,

Thanks for your response.

I do not like the setup but this is how I found it. 

I see " AMQ9407W: Cluster queue <Queue Name> is defined inconsistently"  AND "AMQ9456I: Update not received for object
<Queue Name>, queue manager <QMGR NAME> from full repository for cluster <CLUSTER NAME> " very often in my logs.

Here is another " AMQ9469W: Update not received for CLUSRCVR channel TO.QMGR NAME hosted on queue
manager QMGR NAME in cluster CLUSTER NAME

.

When I looked, I see at CLUSSDR CHL, it has the LOAD BALANCER in the CONNAME, hence my question.

Hope I cleared some misunderstandings and you can advise

Thanks Emile

Hi Emile.

No, it's not recommended to attempt to load balance QM to QM connections of any kind in this way. 

For a client (application) to QMgr connection, load balancing can be viable if correctly configured.  However, QM to QM channel protocols require that connections always reach the exact same QM (as you describe).  This is crucial to confirming message/batch integrity, ordering etc.  therefore you will receive errors if the connection is incorrectly routed.
The only case in which some form of 'redirection' (I won't say 'balancer') would make sense is if this is an HA environment and the two 'Queue Managers' are in fact instances of the exact same queue manager (Multi Instance, RDQM, OS level clustering etc.)
If you can describe what you are trying to achieve with such a configuration I or others here may be able to suggest better alternative approaches?

Regards

Anthony

Is it advisable to have a load balancer name in the Connection Name?

Why am I asking this?

if you have 2 QMGRs on the same host using the same port, how would the load balancer know which QMGR to start the channel on?

What if you have 2 QMGRs on different host with the same port behind the same load balancer?

If so, must one populate the local communication Address as well?