MQSeries.net :: View topic

Sosed · Posted: Tue May 15, 2018 10:26 am Post subject: How MQ Client works with cluster

Hello guys.

I have a question about MQ Client ability to work with clustered queues.

Everybody knows "classic" case: MQ Client connects to a partial repository node (PR1) and writes a message into a cluster queue, which is owned by another pertial repository node (PR2). It is possible because PR1 node gets connection info from one of full repositories and dinamically creates a channel to PR2.

And it works fine even if all FR nodes go down (assuming PR1 has already connected to PR2 at least once):

Here is the question. Is it possible for MQ Client to work the same way as PR1? To dynamically establish a connection to PR2 (without explicit manual configuration, as PR1 in the previous example did)?

For example, MQ Client connects to one of full repositories (FR1) via CCDT. It has no information about other members of the cluster, only connection to FR1 node. Application writes a message via MQ Client into a cluster queue, which is owned by PR2.

Is it possible for MQ Client to communicate with cluster queue if both FR go down?

I didn't find any info supporting this idea.
I also tried to run the case (with rfhutilc and simple Java app (MQ classes for Java) as Clients; I didn't use CCDT, only direct connections) - no success.
So, it's impossible? Or I missed something (like specific options)?

_____________________________________________________________

And a couple of question more:
2. When MQ Client writes into a cluster queue, it works with only one QM (which the client connected to) and the QM redirects messages? Or MQ Client establishes a connection to the target QM?
3. Can MQ Client "cache" cluster topology after connection to one of the cluster members and then use cached info to connect in a random way to any of cluster members without manual configuration like connection list, CCDT?

Thank you in advance!