Clustering vs HA

Hi,

I am trying to understand the difference between clustering and high availability.

In HA, when I have two routers configured as active and standby, when the active jms-router goes down, my JMS clients (senders as well as receivers) automatically switch to the standby.

However, does the same happen in a cluster setup.

Let us say, I have two routers R1 and R2 in my "federated router network", and I have a clustered queue that is using the R1-R2 cluster. Now, let us say, I have a message sender that is sending messages to this clustered queue (see snippet from downloaded sample from SwiftMQ).

                String smqpURL = "smqp://localhost:4001";
                String qcfName = "plainsocket@router2";
                String queueName = "testqueue@router2";

So, when router2 goes down, will I get an exception in my Java code, or will JMS connection automatically switch to router1.

Thanks in advance.

Thanks,
Kartick

[IIT Software]

Kartick Suriamoorthy @ zp wrote:

In HA, when I have two routers configured as active and standby, when the active jms-router goes down, my JMS clients (senders as well as receivers) automatically switch to the standby.

That's correct. HA makes a single router high available by providing a STANDBY instance which takes over when the ACTIVE instance fails. A HA router is a single logical router consisting of 2 phys. HA instances for fault tolerance.

This is independent of a Federated Router Network. You can build any topology with it but only if you make particular nodes (routers) of such a network high available you will have transparent reconnect to the same router (with HA this will be the former STANDBY). You can of course build a Federated Router Network with non-HA routers but if one router goes down, the messages stored at this router will become unavailable until the router is restarted. JMS clients connected to this router will try to reconnect to this router (not to other routers) until max retries has been reached which leads to a JMS exception at the client.

To get over this, build the Federated Router Network or parts thereof with HA routers.

Hi,

thanks for the information. It helped clarify a few things in my mind.

Can you tell me what is maximum number of retries and where is this
configured. Because, I shutdown one of my routers while my client was
sending JMS messages to a clustered queue, and I didn't see any exception.

Thanks,
Kartick

[IIT Software]

It's configured in the connection factory. Look here for SwiftMQ Router Standard and here for SwiftMQ HA Router.

Default reconnect delay is 10 secs and max retry is 50. So after 50 x 10 sec the client should terminated. However, there is a bug fix in 7.5.3 concerning the termination after max retries have been reached.