7 Replies Latest reply on Apr 7, 2009 7:31 AM by ataylor

Clustered example and a few other example things

timfox Apr 6, 2009 2:12 PM

I have taken Howard's simple clustered-queue example and got it to work.

The example programmatically creates two connection factories one to each node.

It then creates a consumer on each node and a producer on one node. It sends 10 messages and verifies that the messages are received round robin.

The example should probably be renamed to ServerLoadBalancing example or something similar.

I had to make a few changes to the config to get this to work.

One thing that I found confusing is the usage of two beans files - one for each server. This meant I had to change all the directories journal, bindings, large messages so each server was using a distinct set.

If we just had a different config directory for each server we wouldn't have to do that.

Also I found that start() was being called on each component twice during the startup - one time by the MC before it injects and another time inside the MessagingServerImpl start method.

We need to add the annotation to prevent the MC calling start/stop, Andy told me what this was the other day but I have forgotten. This still needs to be added.

Other stuff - there is lots of unnecessary cruft in the config files.

1. Re: Clustered example and a few other example things

timfox Apr 6, 2009 2:14 PM (in response to timfox)

For a test that uses JNDI to get the connection factorys and queus then we'll need to create two initial contexts.

I don't know if the example framework currently supports that.
Actions
2. Re: Clustered example and a few other example things

timfox Apr 6, 2009 3:12 PM (in response to timfox)

I also added a clustered topic example.

Note that I didn't update the readme etc - that still needs to be done but it gives an idea of how to put these clustering examples together.
Actions
3. Re: Clustered example and a few other example things

gaohoward Apr 6, 2009 10:03 PM (in response to timfox)

Thanks Tim.

>>>For a test that uses JNDI to get the connection factorys and queus then we'll need to create two initial contexts.

I think we can configure JNDI on each means config file using different ports. But there is only one client-jndi.properties, so i think for clustered example, we have to hard-code the jndi ENV.
Actions
4. Re: Clustered example and a few other example things

gaohoward Apr 6, 2009 11:26 PM (in response to timfox)

Hi Tim,

I think using JNDI for connection factory and queue lookup in clustered examples is more consistent with other jms examples. We can show more of other ways to do this in the core examples.
Actions
5. Re: Clustered example and a few other example things

ataylor Apr 7, 2009 5:44 AM (in response to timfox)

One thing that I found confusing is the usage of two beans files - one for each server.

We need 2 beans configs as the naming server ports are configured in them. I could still create 2 configs dirs to make things simpler but each would have its own beans file and configuration.
Actions
6. Re: Clustered example and a few other example things

ataylor Apr 7, 2009 6:10 AM (in response to timfox)

This meant I had to change all the directories journal, bindings, large messages so each server was using a distinct set.

You will have to do this anyway since the they are created relatively from the example base dir not the config dir.
Actions
7. Re: Clustered example and a few other example things

ataylor Apr 7, 2009 7:31 AM (in response to timfox)

Ive made the examples a lot simpler to write.

each server now has its own config, this will be in server0, server1 etc.

there should also be a client-jndi.properties in each config. Then in the code you can call getContext(id) where id is the server you want to connect too.
Actions

Go to original post