[jboss-jira] [JBoss JIRA] (WFLY-11198) Session draining on stopped worker won't complete if further requests are made on balancer

[Jan Kašík (Jira)]

     [ https://issues.jboss.org/browse/WFLY-11198?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jan Kašík updated WFLY-11198:
-----------------------------
                Tester: Jan Kašík
    Steps to Reproduce: 
# Start mod_cluster configuration with at least two workers (with session-draining-strategy=ALWAYS) and one balancer
# Deploy simple application with simple {{web.xml}} defining session timeout: {code:xml}
<web-app>
    <distributable />
    <session-config>
        <session-timeout>1</session-timeout> <!-- minutes -->
    </session-config>
</web-app>
{code} on each worker
# Make request on balancer to context created by deployed application and check which worker handled this request
# Disable this worker {{/subsystem=modcluster/proxy=default:stop(waittime=0)}}
# Check, that contexts on this worker are {{disabled}}
# Make further requests on balancer to the same context
# Draining will never complete


> Session draining on stopped worker won't complete if further requests are made on balancer
> ------------------------------------------------------------------------------------------
>
>                 Key: WFLY-11198
>                 URL: https://issues.jboss.org/browse/WFLY-11198
>             Project: WildFly
>          Issue Type: Bug
>          Components: mod_cluster
>    Affects Versions: 15.0.0.Alpha1
>            Reporter: Jan Kašík
>            Assignee: Radoslav Husar
>            Priority: Major
>
> When disabling worker with {{/subsystem=modcluster/proxy=default:stop(waittime=0)}}, the contexts are switched to {{disabled}} state. If there no further requests (on the same context which session currently times out) are made on a balancer, session correctly times out, contexts switches to {{stopped}} state and everything is OK.
> However, when I run same stop operation, wait until contexts are switched into disabled state and in time in which sessions time out I make more requests on balancer, contexts will never switch to {{stopped}} state and session draining never completes. This happens intermittently (~60% of cases with 10 requests made after {{disabled}} state is in place).



--
This message was sent by Atlassian Jira
(v7.12.1#712002)