Ad Widget

Collapse

How to monitor many machines on each running 2 Tomcat instances using JMX?

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • swachter
    Junior Member
    • Feb 2013
    • 11

    #1

    How to monitor many machines on each running 2 Tomcat instances using JMX?

    Hi all,

    we plan to monitor approx. 50 servers, where on each server 2 Tomcat instances are running. We would like to use JMX to get the values periodically.

    I assume, that we should have an agent on each machine. Yet, I did not find a clue in the documentation that agents can use the Java gateway. Do we need some custom code or can this be done be configuration?

    Can someone tell what the proper setup should look like?

    TIA
    Stefan
  • Slash
    Member
    • May 2011
    • 64

    #2
    If you only want to monitor tomcat, you don't need any agent on the servers.

    The zabbix java gateway runs on a different server (can be the same as the zabbix server) and it connects itself to the tomcat you want to monitor using jmx beans (native JVM monitoring, like jconsole for example).

    You only need to enable remote monitoring on your tomcats (protected by a login/password, preferably) and you'll be all set.

    The only downside is that since you have 2 different tomcat, you'll have to set different "management.jmxremote.port" on each one and adapt your template accordingly.

    Comment

    • swachter
      Junior Member
      • Feb 2013
      • 11

      #3
      Hi Slash,

      thanks for your reply. Ok, I understand we need to install one Zabbix Java Gateway.

      How many StartJavaPollers and START_POLLERS should be configured? There are approx. 50 x 2 = 100 Tomcat instances to be monitored. How does Zabbix server schedule the read operations? Does Zabbix server try to monitor all instances at the same time? In that case there should be 100 pollers configured each on Zabbix Server and Zabbix Java Gateway. If Zabbix server distributes the monitoring of all instances equally over the monitoring intervall however, then only a few pollers should suffice.

      PS: Monitoring the instances at the same time would be preferable because then all read values reflect the state of the same instance in time.

      Comment

      Working...