Created attachment 917693 [details] Reproducer deployment and log files Description of problem: On deploying an EJB which has wrong "lookup" JNDI names, it gives ERROR log on deployment, but if you restart EAP then, failed deployment becomes "deployed" and there is only INFO log on startup. Tested this on both EAP 6.2.0 and 6.2.4. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1) Start up vanilla installed EAP first(standalone-full.xml), drop the folder dsc.ear into $JBOSS_HOME/standalone/deployments, then create an empty file "dsc.ear.dodeploy" 2) EAP gives ERROR log (see attachment "LogOnDeployment.log"), and deployment failed ("dsc.ear.dodeploy"==>"dsc.ear.failed"), cannot see this deployment in web console (http://localhost:9990) 3) Now restart EAP, it gives INFO log now(see attachment "LogOnRestart.log"), and deployment succeeded ("dsc.ear.failed"==>"dsc.ear.deployed"), can see this deployment in web console Actual results: The deployment statuses are not consistent before/after the restart. On restart there should be ERROR level logging followed by some stacktrace. Expected results: Deployment status should remain "failed" and shouldn't be visible in web console after server restart. On restart the logging of "JBAS014775: New missing/unsatisfied dependencies:" should be in ERROR level and some stacktrace is preferred. Additional info:
Source code see "TestJMX.java" and "TestJMXMBean.java" in the attachment, we use an invalid JNDI name to "lookup" the @Resource [lookup="java:global:/" +] , which should be [lookup="java:global/" +] @Resources //EAP6 ( { @Resource(name="java:jboss/" + TestJMXMBean.JNDI_NAME, lookup="java:global:/" + "dsc" + '/' + "tpapps-common-testdeploy" + '/' + "TestJMX" + '!' + TestJMXMBean.JNDI_NAME ) } )
Moreover, if using standalone-full-ha.xml to restart, it doesn't even give anything in the log, all the log entries look perfect (and saying the ear is deployed successfully), the only thing indicating an error is the last sentence in the server startup logging: ============================== 11:32:30,319 ERROR [org.jboss.as] (Controller Boot Thread) JBAS015875: JBoss EAP 6.2.4.GA (AS 7.3.4.Final-redhat-1) started (with errors) in 4202ms - Started 230 of 365 services (5 services failed or missing dependencies, 129 services are passive or on-demand) ==============================
This issue can not be reproduced with EAP-6.4.0.DR8. It is probably same problem as https://bugzilla.redhat.com/show_bug.cgi?id=997583 and is already fixed in EAP-6.4.0.DR1. *** This bug has been marked as a duplicate of bug 997583 ***
"java:global:/" - why extra ':' ? -- Just a typo. Customer found this issue originally with this typo in their code. And they spent a lot of time troubleshooting this as there is no clear pointer for the actual problem. -------------------------------------------------------- Ref: - https://issues.jboss.org/browse/WFLY-3623 - https://bugzilla.redhat.com/show_bug.cgi?id=1119531 So I guess its still a duplicate. -- It's different, bz 1119531 is about "Binding JNDI names incorrectly" rather than bz 1119159's "Not reporting missing/invalid JNDI references" If we try to use "java:jboss:/ResourceTestEJB" to bind the JNDI name, it will bind "java:jboss:" instead and there will be no error (actually we expect to see an error saying invalid name/namespace etc.) Please check bz 1119531: -------------------------------------------- An @Resource annotation with a spurious ":" in the name attribute value results in an incorrect binding but no deployment failure. @Resources({ @Resource( // name="java:jboss/ResourceTestEJB", // ==> this works fine name="java:jboss:/ResourceTestEJB", // ==> this will not bind the JNDI correctly, and there is no error on deploying EJB ...... Actual results: "java:jboss:" is bound as JNDI name and the resource cannot be looked up. Expected results: If a proper binding can't be done per the annotation value, there should be an exception thrown triggering deployment rollback. --------------------------------------------
bz997583 is about "file-scanner deployment that fails on boot will not be undeployed after failure" and incorrectly change deployment statue from .failed marker to .deployed marker after server reboot, so, it's not fix for any specific reason that how make it fails(unlike this "wrong 'lookup' JNDI names" it could be just missing dependency in jboss-deployment-structure.xml or other situation)
Brian Stansberry <brian.stansberry> updated the status of jira WFLY-3625 to Resolved