[
https://issues.redhat.com/browse/WFLY-13177?page=com.atlassian.jira.plugi...
]
Guido Jäkel updated WFLY-13177:
-------------------------------
Description:
On WF-13 and WF-16 we observe a serious bug of the RequestCount of the RequestController
while using ManagedExecutorService.submit() or ...execute() in the edge case of a full
queue. In this case, the caller gets a RejectedExecutionException, but in the
RequestController, the number of active requests is erroneously incremented.
This will lead to a false and monotonously increasing number of active requests. And in
case of a limitation configured by the maxRequestCount feature, which is best practice for
production environments, over the time this will lead to deadlock of the RequestController
and herewith the complete activity of the Wildfly at all.
was:
On WF-10 and WF-16 we observe a serious bug of the RequestCount of the RequestController
while using ManagedExecutorService.submit() or ...execute() in the edge case of a full
queue. In this case, the caller gets a RejectedExecutionException, but in the
RequestController, the number of active requests is erroneously incremented.
This will lead to a false and monotonously increasing number of active requests. And in
case of a limitation configured by the maxRequestCount feature, which is best practice for
production environments, over the time this will lead to deadlock of the RequestController
and herewith the complete activity of the Wildfly at all.
ManagedExecutorService: Wrong activeRequestCount at RequestController
on RejectedExecutionException
---------------------------------------------------------------------------------------------------
Key: WFLY-13177
URL:
https://issues.redhat.com/browse/WFLY-13177
Project: WildFly
Issue Type: Bug
Components: EE, Server
Affects Versions: 13.0.0.Final, 16.0.0.Final
Reporter: Guido Jäkel
Assignee: Brian Stansberry
Priority: Major
On WF-13 and WF-16 we observe a serious bug of the RequestCount of the RequestController
while using ManagedExecutorService.submit() or ...execute() in the edge case of a full
queue. In this case, the caller gets a RejectedExecutionException, but in the
RequestController, the number of active requests is erroneously incremented.
This will lead to a false and monotonously increasing number of active requests. And in
case of a limitation configured by the maxRequestCount feature, which is best practice for
production environments, over the time this will lead to deadlock of the RequestController
and herewith the complete activity of the Wildfly at all.
--
This message was sent by Atlassian Jira
(v7.13.8#713008)