[arquillian-issues] [JBoss JIRA] (ARQ-1949) Memory leak in DeploymentContextImpl.stores

[Falko M. (JIRA)]

    [ https://issues.jboss.org/browse/ARQ-1949?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13066427#comment-13066427 ] 

Falko M. commented on ARQ-1949:
-------------------------------

A workaround is to call destroy explicitly:
{code:java}
public class LeakFixHandler {

    @Inject
    private Instance<DeploymentContext> deploymentContext;

    public void destroyDeployment(@Observes(precedence = Integer.MIN_VALUE) final UnDeployDeployment event) {
        deploymentContext.get().destroy(event.getDeployment());
    }
}
{code}

> Memory leak in DeploymentContextImpl.stores
> -------------------------------------------
>
>                 Key: ARQ-1949
>                 URL: https://issues.jboss.org/browse/ARQ-1949
>             Project: Arquillian
>          Issue Type: Bug
>    Affects Versions: 1.1.3.Final, 1.1.7.Final
>         Environment: Maven 3.2.5
> Surefire-Plugin 2.18.1
> Windows7 x64 and CentOS 6.5
>            Reporter: Falko M.
>
> The {{Deployment}} (which also holds the {{Archive}}) for each test class is added to {{DeploymentContextImpl.stores}}. A {{Deployment}} is deactivated properly but it is *not* getting removed until the {{Manager}} shuts down.
> This can easily blow the heap of the executing JVM in case you have *many* test classes and/or *large* archives to deploy.
> In our environment the JVM forked by maven surefire dies with an OOM after just ~20 test classes. Xmx is set to 512m.
> PS: Another user came across this problem too: https://developer.jboss.org/thread/250124



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)