If a not clustered EJB application is deployed and a client try to access it by defining only one server connection the expectation is that exact this server is used and, if this server failed, the EJB can not longer be invoked. In case of a clustered server environment this will be the case as long as there is no clustered application and the cluster subsystem is not started. But if there is a clustered application deployed at the same time the non clustered application change the behaviour to soemwhat clustered with LB and failover. 1) it is not expected that a non clustered application has some kind of LB and failover without any configuration at client side as it is intentional that there is only one server configured and the application is not clustered! 2) The behaviour of application / client should not effected by other deployed applications 3) The credentials are not passed to the servers which are added internaly if the configuration does not contain the "cluster" part The same problem is visible if there is a mix of clustered and non clustered EJB's in one application archive
Proposing for EAP 6.4.
This is just a side effect of the design. If you don't like this behavior you have three options: 1) Use distinct name 2) Don't deploy EJBs on servers you don't want clients to call 3) Write your own DeploymentNodeSelector, and use deployment.node.selector
> This is just a side effect of the design. Then the design is broken and needs to be fixed. If the behavior of one application changes simply because another unrelated application is deployed on the same server, that's a bug.
The definition of a bug is behavior that is not intended.
Jason Greene <jason.greene> updated the status of jira EJBCLIENT-102 to Closed
Jason Greene <jason.greene> updated the status of jira EJBCLIENT-102 to Reopened
Just to be clear though, we are planning on moving to a new invocation behavior pattern. We just can't really fix this in a 6.x release. The best that could be done is a workaround of some kind, but we are swamped and can't do everything.
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days