Details
-
Bug
-
Resolution: Done
-
Major
-
16.0.0.Beta1
-
None
Description
The reproducer shows a sequence of operations where a server group is completely removed and re-added later. If the server group has a server that is using its own jvm definition, when the server and server group are removed and added again, the system properties that the server group had are still shown in the internal model managed by the DC.
This situation leads to an inconsistent state between the model persisted and the runtime model managed by the DC. The persisted model on the DC configuration (domain.xml) doesn't contain the system property at the server group, but the model managed at runtime has such an obsolete property. Read operations don't show this inconsistency.
At this point, if we try to re-add the same system property that the server group had, we get the following error:
[domain@localhost:9990 /] /server-group=main-server-group/system-property=test:read-resource { "outcome" => "failed", "failure-description" => "WFLYCTL0216: Management resource '[ (\"server-group\" => \"main-server-group\"), (\"system-property\" => \"test\") ]' not found", "rolled-back" => true } [domain@localhost:9990 /] /server-group=main-server-group/system-property=test:add(value=true) { "outcome" => "failed", "failure-description" => {"host-failure-descriptions" => {"slave" => "WFLYCTL0212: Duplicate resource [ (\"server-group\" => \"main-server-group\"), (\"system-property\" => \"test\") ]"}}, "rolled-back" => true, "result" => {} }
Another way to verify this inconsistency is to start the server and inspect the java options passed by the HC. The obsolete property is still being used:
ps -fea | grep java | grep server-one | grep test