I am working on Open High Availability Cluster, the open source version of Solaris Cluster, that makes applications HA. Glassfish is one of those applications.
While this was working fine from Application Server 7.0 till glassfishv2, glassfishv2ur2 has a few features that are not working the way it used to.
In cluster configuration, DAS and Node agents are made highly available. DAS is made to listen
on a virtual or logical IP that is used to failover from one machine to another when the app becomes unavailable. Similar is the case with Node agents to listen on another logical IP.
In order to achieve this, the HTTP_LISTENER of das and node agents will be changed in domain.xml, das.properties and nodeagent.properties to listen on a virtual IP that gets umplumbed on one machine and plumbed on to the other machine. Typical configuration being, 1 DAS, 2 Node agents, 4 instances under each node agent and 2 clusters.
Issue 1: After making changes to domain.xml to listen on logical IP address, DAS takes longer time(>20minutes) to come up if the logical IP on which node agent is made to listen is not alive/running. This was not the case earlier. Due to this behavior, our agent timesout and tries to restart DAS continuously, which will not be desirable to the Application server HA agent customers. In a single standalone machine, everything works fine.
Here is a snip of server.log:
[#|2008-07-21T12:37:52.861+0530|INFO|sun-appserver9.1|javax.enterprise.system.core.transaction|_ThreadID=10;_ThreadName=main;3700;|JTS5014: Recoverable JTS instance, serverId = [3700]|#]
[#|2008-07-21T*[b]12:37:53[/b]*.645+0530|INFO|sun-appserver9.1|javax.enterprise.system.tools.admin|_ThreadID=10;_ThreadName=main;|ADM1079: Initialization of AMX MBeans started|#]
[#|2008-07-21T*[b]12:52:52*.96[/b]1+0530|INFO|sun-appserver9.1|javax.enterprise.system.tools.admin|_ThreadID=10;_ThreadName=main;|amx:j2eeType=X-J2EECluster,name=CL1 is registered.|#]
[#|2008-07-21T12:52:52.964+0530|INFO|sun-appserver9.1|javax.enterprise.system.tools.admin|_ThreadID=10;_ThreadName=main;|amx:j2eeType=X-J2EECluster,name=CL2 is registered.|#]
Note the timestamp in bold. It takes longer time to come to the next stage of registering the clusters. Note that this happens only in the (app server) cluster configuration.
Issue 2: jmx-connector of Cluster-config tag now has the client-hostname property tag. (pphysical-2 is the physical hostname)
<property name="client-hostname" value="pphysical-2"/>
It was present only for domain and node agents in the previous releases. Is there a change from this release onwards?
Can somebody throw some light on this?
Thanks!
[Message sent by forum member 'sdevulap' (sdevulap)]
http://forums.java.net/jive/thread.jspa?messageID=292249