CLI: creating a server on a domain slave followed by setting a system property fails when done in batch mode ------------------------------------------------------------------------------------------------------------ Key: WFLY-1309 URL: https://issues.jboss.org/browse/WFLY-1309 Project: WildFly Issue Type: Bug Components: Domain Management Affects Versions: 8.0.0.Alpha1 Environment: wildfly build from a git clone on 2013-05-02 Reporter: Tom Fonteyne Assignee: Brian Stansberry Fix For: 8.0.0.CR1 Attachments: host-domain-controller.xml, host-slave.xml Setup two boxes: - on one, run domain mode (full-ha) -> domain controller - on the other, configure host.xml to point to the DC, and run the host controller Now in CLI: # add the group: /server-group=slaves:add(profile=full-ha,socket-binding-group=full-ha-sockets) # create server and set a prop batch /host=slave1/server-config=i1:add(group=slaves,auto-start=false) /host=slave1/server-config=i1/system-property=env:add(value=dev) run-batch results in: JBAS010839: Operation failed or was rolled back on all servers. Debugging (on EAP 6.0.1) showed the real reason: activeStep.response = (org.jboss.dmr.ModelNode) { "outcome" => "failed", "failure-description" => "JBAS010850: No handler for operation composite at address [ (\"host\" => \"slave1\"), (\"server\" => \"i1\") ]", "rolled-back" => true } More debugging led to : org.jboss.as.controller.AbstractOperationContext seems to be where the rollback is set. When debugging, I found that the constructor always sets ModelController.OperationTransactionControl transactionControl but at line 332 (eap 6.0.1) / line 372 (wildfly), the end of doCompleteStep() : // Allow any containing TransactionControl to vote final AtomicReference<ResultAction> ref = new AtomicReference<ResultAction>(transactionControl == null ? ResultAction.KEEP : ResultAction.ROLLBACK); and ref becomes a rollback. I "walked" pretty much of all the code going on in between. The bit where the config is getting persisted is successful. But I'm afraid I'm out of my depth as there is *a lot* going on to execute such a composite command :)