Created attachment 828657 [details] stacktrace and standard output See e.g.: https://jenkins.mw.lab.eng.bos.redhat.com/hudson/view/EAP6/view/EAP6-AS-Testsuite/job/eap-60-as-testsuite-RHEL-matrix-OracleJDK7/98/jdk=java17_default,label_exp=RHEL6%26%26x86_64/testReport/org.jboss.as.test.manualmode.ws/ReloadWSDLPublisherTestCase/testHelloStringAfterReload/
The issue is due to the fact that the jvm is reusing the same connection. Setting keep-alive to false fixes it.
Emmanuel is right, but keep-alive cannot be used with WS invocations (you don't have access to the underlying connection), so that would only resolve half of the problem. I think the best solution will be to add a 5 second wait until the pooled connections time out.
It is working by setting the keep-alive to false in the set up and resetting it to the previous value in the tear down. I have a patch for this. This avoid long( 5s) test.
https://github.com/ehsavoie/jboss-eap/tree/BZ-1034229 : the patch but I need to make one for upstream.
Ok, I didn't think of this solution. We'll go with yours then.
PR : https://github.com/jbossas/jboss-eap/pull/720 Upstream : https://github.com/wildfly/wildfly/pull/5576
Fixed for both ipv4 and dualstack in 6.3.0 DR0. I filled new https://bugzilla.redhat.com/show_bug.cgi?id=1067348 for pure ipv6.