Do you mean manually monitor (like a browser page), or automatically (like alerts go out when stuff doesn't appear to be running)?
We use Nagios (formerly Netsaint) to automatically monitor availability of various services in our cluster. It has proved a very useful tool. Since it can do URL checks, you can wrap more "functional" checks into a servlet or JSP and point a Nagios service to the URL of that check. We even use a JSP to iterate over the datasources boundin JNDI and do a basic functional check. Also, you can use basic socket checks from Nagios to ensure availability of listening services (JNDI, RMI, etc).
For manual checks, jmx-console and web-console.
Thank You Mike. It is useful. But can you show me the steps of how to use JMX console to monitor ? Actually, I just want to do something with JMX to monitor the clusters of JBoss.
Sure. Can you be more specific in the things you want to monitor? The cluster service itself? Clustered applications? Have you looked at the jmx-console and web-console webapps that are pkgd with the server?
um...I need to monitor every things happen in the cluster, includes application and the JBoss server itself. So I am trying to find out how JBoss clusters node communicate with each other. So I want to know how to use JMX-Console to check the cluster's status.
I have just started a JBoss cluster on two windows machines. At least, I think so. JBoss runs in the "all" configuration and 'loopback' has been set to 'true' in cluster-service.xml on both machines. As I undersytand it, both nodes should have discovered each other. Is there an easy way to check this? Can you ask a JBoss cluster which nodes are participating in the cluster?
Start node 1 and let the start up finish.
Start node 2, if the cluster is formed correctly you will see an obvious message on node 1 showing that the membership of the cluster has changed.
Thanks, I see now: at startup, JBoss will write informational messages regarding clustering. I can find them if I look for the word "ClusterPartition" in the log file (JBoss system output is written to a log file in my case).
Also nice is that in the same log file messages are written about the number of dead cluster members. Apparently, JBoss creates an informational message each time the number of nodes in a cluster changes.
I guess I did not see this before because clustering was not configured right yet.
I am curious: is it possible to identify dead cluster members from the jmx- or adminstration console?
Like Arthuray, I am looking for a way to monitor or even manage a JBoss cluster. Unfortunately, Nagios does not run on Windows OSs.
So, is there any monitor tool for windows?
I believe the Windows OS itself has something called WMI that you might be able to use.