[JBoss JIRA] (MODCLUSTER-711) Using "connectorPort" property fails if multiple services are configured in Tomcat
by Tomas Briceno Fernandez (Jira)
[ https://issues.redhat.com/browse/MODCLUSTER-711?page=com.atlassian.jira.p... ]
Tomas Briceno Fernandez commented on MODCLUSTER-711:
----------------------------------------------------
[~rhusar] They have separate connectors for managment and applications, I think this is the reason why they create separate services. I'm not sure if multi-engine set up can accommodate that.
As per mod_cluster documentation, it is expected that either *connectorPort* or *connectorAddress* is set when configuring on Tomcat 9. Therefore the expectation is that the service matching that setting must be the one registered.
> Using "connectorPort" property fails if multiple services are configured in Tomcat
> ----------------------------------------------------------------------------------
>
> Key: MODCLUSTER-711
> URL: https://issues.redhat.com/browse/MODCLUSTER-711
> Project: mod_cluster
> Issue Type: Bug
> Components: Core & Container Integration (Java)
> Affects Versions: 1.4.0.Final
> Reporter: Tomas Briceno Fernandez
> Assignee: Radoslav Husar
> Priority: Major
> Fix For: 2.0.0.Alpha1, 1.4.2.Final
>
>
> If the Tomcat server configuration has several <service> elements and the mod_cluster listener is configured with *connectorPort* (most probably it is the same with *connectorAddress*), the configuration fails with these messages:
> {code}
> 06-Feb-2020 16:11:17.596 INFO [ContainerBackgroundProcessor[StandardEngine[TestEngine]]] org.jboss.modcluster.ModClusterService.connectionEstablished MODCLUSTER000012: TestEngine connector will use /127.0.0.1
> 06-Feb-2020 16:11:17.598 INFO [ContainerBackgroundProcessor[StandardEngine[TestEngine]]] org.jboss.modcluster.ModClusterService.establishJvmRoute MODCLUSTER000011: TestEngine will use 7bb39e02-96c0-3f8f-9fab-d464ad729cfe as jvm-route
> 06-Feb-2020 16:11:17.598 SEVERE [ContainerBackgroundProcessor[StandardEngine[TestEngine]]] org.apache.catalina.core.ContainerBase$ContainerBackgroundProcessor.processChildren Exception invoking periodic operation:
> java.lang.RuntimeException: MODCLUSTER000047: No configured connector matches specified host:port (*:8081)! Ensure connectorPort and/or connectorAddress are configured.
> at org.jboss.modcluster.container.tomcat.ConfigurableProxyConnectorProvider.createProxyConnector(ConfigurableProxyConnectorProvider.java:89)
> at org.jboss.modcluster.container.tomcat.TomcatEngine.getProxyConnector(TomcatEngine.java:140)
> at org.jboss.modcluster.ModClusterService.connectionEstablished(ModClusterService.java:267)
> at org.jboss.modcluster.mcmp.impl.DefaultMCMPHandler.status(DefaultMCMPHandler.java:341)
> at org.jboss.modcluster.mcmp.impl.DefaultMCMPHandler.status(DefaultMCMPHandler.java:315)
> at org.jboss.modcluster.ModClusterService.status(ModClusterService.java:388)
> at org.jboss.modcluster.container.tomcat.TomcatEventHandlerAdapter.lifecycleEvent(TomcatEventHandlerAdapter.java:229)
> at org.apache.catalina.util.LifecycleBase.fireLifecycleEvent(LifecycleBase.java:123)
> at org.apache.catalina.core.ContainerBase.backgroundProcess(ContainerBase.java:1174)
> at org.apache.catalina.core.ContainerBase$ContainerBackgroundProcessor.processChildren(ContainerBase.java:1396)
> at org.apache.catalina.core.ContainerBase$ContainerBackgroundProcessor.run(ContainerBase.java:1368)
> at java.lang.Thread.run(Thread.java:748)
> {code}
> My own code inspection suggests this is because of this loop :
> {code:title=org.jboss.modcluster.ModClusterService}
> @Override
> public void connectionEstablished(InetAddress localAddress) {
> for (Engine engine : this.server.getEngines()) {
> Connector connector = engine.getProxyConnector();
> InetAddress address = connector.getAddress();
> // Set connector address
> if ((address == null) || address.isAnyLocalAddress()) {
> connector.setAddress(localAddress);
> ModClusterLogger.LOGGER.detectConnectorAddress(engine, localAddress);
> }
> this.establishJvmRoute(engine);
> }
> this.established = true;
> }
> {code}
> The problem here is that the invocation of *engine.getProxyConnector()* will check if one and only one of the connectors in the engine matches the port configured by *connectorPort*. If more that one service is configured there will be multiple engines and this code will apply the previous condition to all of them. That is, to properly exit this method the port should exist in all engines, which will not happen normally.
--
This message was sent by Atlassian Jira
(v7.13.8#713008)
4 years, 9 months
- 1
- 0
[JBoss JIRA] (MODCLUSTER-703) Handle multiple JSESSIONID with overlapping domains or paths
by Radoslav Husar (Jira)
[ https://issues.redhat.com/browse/MODCLUSTER-703?page=com.atlassian.jira.p... ]
Radoslav Husar commented on MODCLUSTER-703:
-------------------------------------------
[~fcornuau] BTW have you by any chance tested how this works with the Java implementation when using WildFly as a mod_cluster load balancer? https://docs.wildfly.org/18/Admin_Guide.html#Undertow_using_as_a_Load_Bal...
> Handle multiple JSESSIONID with overlapping domains or paths
> -------------------------------------------------------------
>
> Key: MODCLUSTER-703
> URL: https://issues.redhat.com/browse/MODCLUSTER-703
> Project: mod_cluster
> Issue Type: Feature Request
> Components: Native (httpd modules)
> Affects Versions: 1.3.8.Final
> Environment: Apache 2.4.x
> jbcs-httpd24-mod_cluster-native-1.3.8-3.Final_redhat_2.jbcs.el7.x86_64
> Reporter: Frederic CORNUAU
> Assignee: Jean-Frederic Clere
> Priority: Critical
> Labels: JSESSIONID, mod_cluster
>
> We currently migrate our Apache LB mechanism from _mod_jk_ to _mod_proxy_cluster_ and we face an issue when multiple JSESSIONID (with overlapping domains or paths) are being received.
> mod_jk supports multiple JSESSIONID cookies and is able to select the most suitable one based on the jvmRoute (see https://bz.apache.org/bugzilla/show_bug.cgi?id=44116 and https://github.com/apache/tomcat-connectors/blob/trunk/native/common/jk_l...).
> On the other hand, _mod_proxy_cluster_ uses only the first JSESSIONID received in the "Cookie" HTTP request header (see https://github.com/modcluster/mod_cluster/blob/1.3.8.Final/native/mod_pro...).
> Is-it possible to update _mod_proxy_cluster_ implementation to handle multiple JSESSIONID cookies?
--
This message was sent by Atlassian Jira
(v7.13.8#713008)
4 years, 9 months
- 1
- 0