mfarver's questions

We have a network of HP 2510G switches connected back to HP2912al for aggregation. We've noticed that long running connections like a MySQL DB dump start flooding out to all network ports once the mac-cache-timeout expires. Doing an "arping" against the destination IP stops the flooding (going back to port to port) until the cache timeout expires again.

I can understand why this would happen for unidirectional UDP traffic, but I'm at a loss as to why it is happening for TCP. I would think the ACKs from the receiving machine would cause the Procurves to refresh the MAC address in their cache. Instead it seems like they only learn from ARPs.

Any ideas?

Anyone know of a way to "boot" another Linux kernel from within Linux, discarding the original kernel/initrd? The idea is that I can boot my server from a small read only copy of Linux that does some housekeeping/data gathering tasks and then checks in with a management server to figure out what to do with the server. Most of the time that will be to just boot Linux/Windows off the local harddrives or image the box.

Basically, I want a tiny copy of Linux to be the bootloader, instead of a specialized (and less flexible) bootloader package. The tiny copy of Linux could be on a USB key, or delivered by PXE.

Loadlin used to do this for Windows, so I could create this environment in DOS... but networking can be tricky and I lose a lot of tools.

Other options are virtualization. All the automation above is part of the host OS, and it checks in with the management server to figure out guests to launch and the like. How the cloud servers do it. Or I could do some magic on the PXE side: PXE boot the data gathering copy of Linux first, make a change to the PXE server and reboot it with the second time it PXE booting a remote kernel but a local OS drive.

I am trying to get an application working on a JBoss Cluster. It uses Queues internally, and the developer claims that it should work correctly in a clustered environment. I have jbossmq setup as a ha-singleton on the cluster. The application works correctly on whichever node currently is running the queue, but fails on the other nodes with a: "javax.naming.NameNotFoundException: QueueConnectionFactory not bound" error. I can look at JNDIview from the jmx-console and see that indeed the QueueConnectionFactory class only appears on the primary node in the Global context. Is there a way to see the Cluster's JNDI listing instead of each server?

The steps I took from a default Jboss 4.2.3.GA installation were to use the "all" configuration. Then removed /server/all/deploy/hsqldb-ds.xml and /deploy-hasingleton/jms/hsqldb-jdbc2-service.xml, copying the example/jms/mysq-jdbc2-service.xml file into its place (editing that file to use DefaultDS instead of MySqlDS). Finally I created a mysql-ds.xml file in the deploy directory pointing "DefaultDS" at an empty database. I created a -services.xml file in the deploy directory with the queue definition. like the one below:

<server>
        <mbean code="org.jboss.mq.server.jmx.Queue"
                name="jboss.mq.destination:service=Queue,name=myfirstqueue">
                <depends optional-attribute-name="DestinationManager">
                        jboss.mq:service=DestinationManager
                </depends>
        </mbean>
</server>

All of the other cluster features of working, the servers list each other in the view, and sessions are replicating back and forth. The JBoss documentation is somewhat light in this area, is there another setting I might have missed? Or is this likely to be a code issue (is there different code to do a JNDI lookup in a clusted environment?)

Thanks