JBossDeveloper

Apart from UDP being somewhat "chatty" on the network, the network admins that I have worked with do not like UDP traffic because they claim it could enable the introduction of a "rogue" server into the cluster, which could then do malicious things.  At one client, someone actually (accidentally) joined the live production cluster by starting JBoss locally on their developer workstation.

IMHO thats a terrible misconfiguration. Few things on top of my mind that just went wrong:

• TTL for UDP packets to be 1 (1 hop, same network, noone else gets messages + performance)
• the switch should never send UDP outside of its network
• unchanged partition name (-g switch)
• no JGroups auth used (see org.jgroups.auth)
• JGroups ENCRYPT protocol can be used to secure communication

Anyway...

Bare with us as we add documentation around JNDI.

Take a look at the Overiew section in particular here:

https://docs.jboss.org/author/display/AS71/JNDI+Reference

In a nutshell, we are recommending everyone move to "Client JNDI" where possible so that it eliminates the unnessecary performance overhead of  extra network hops and removes the single point of failure that a traditional remote jndi server has. We provided support for Traditional Remote JNDI since it was more familar and some want the notion of a customizable central directory. However, we do not cluster it in any way at this time. That said, we might, if there is enough demand add some kind of limited failover cability.

Ed, I realize that in your case you are using it because we do not offer an equivalent "Client JNDI" for hornetq, this is something we plan to contribute back to the HQ project. Until that time, if you really need a clustered client I highly recommend just using the HQ JMS APIs directly instead of JNDI.

How are you modifying them to work with jboss properties and classes? Do they let you provide an InitialContext? Or is it a properties map? Is it all in configuration files or in Java?

Ed Keen wrote:

 

Jaikiran, have you had a chance to discuss this feature with the rest of the team yet? 

I actually forgot about this one and will take it up this week. Thanks for reminding.

Jason Greene wrote:

 

That said, we might, if there is enough demand add some kind of limited failover cability.

 

Ed, since Jason did mention that we might add some kind of feature around this depending on the demand, could you please file a feature request JIRA. Make sure you don't project it as a clustering requirement/feature so that it isn't confusing.

I am also trying to configure Clustered JMS Destination where we can specify multiple provider URL from client side.

Will it work if I map multiple host to one DNS name something like (myclusteraddress.com) and use Provider URL like remote://myclusteraddress.com:4447?

Ravi Singhal wrote:

 

I am also trying to configure Clustered JMS Destination where we can specify multiple provider URL from client side.

 

Just to re-iterate and avoid any kind of confusion - this has nothing to do with clustering.

Hi,

Is this issue fixed?

I have updated my JBOSS from AS 7.1 to EAP 6.2.

This issue still exists there.