Debugging high agent CPU usage when monitoring ESXi hosts using VI3
Hi, We are monitoring 8x ESX 4i hosts using Hyperic agent/server 4.1.2 and configuring the hosts to be monitored by the VI3 interface. This is working reasonably well apart from the fact the CPU taken by the agent process is excessive. We are seeing overall CPU usage of about 80% just for the agent doing the ESX4i monitoring. How can we diagnose on why the vim-plugin.jar is using so much CPU to assist the Hyperic development team?
Hi, incase it helps anyone else... I switched the agent and the server to use JDK 1.6 update 14 and I've noticed a substantial drop in CPU for the Hyperic vi3 ESX agent process. When I look at the release notes on the SUN site between 1.5.0 update 12 (which is shipped with HQ 4.1.2) and 1.6 update 14 there are a number of performance improvements and HTTPS stability fixes which might help to explain why monitoring of the ESXi servers was previously so high. I haven't picked up any issues or errors in the log files to indicate any other issues with running the 1.6 update 14 release instead of the shipped one.
CPU on the monitoring server dropped from arpund 80-90% to an averyage of less than 50% afterwards.
thank you for your comment. Do you run HQ Agent / Server on the same host ? Does the upgrade of the Agent make the real difference ? If it is really an SSL issue, you could try to setup stunnel and use http:// in Configuration Properties of your HQ Agent.
Hi, sorry for the delay in responding - I have been busy on other projects. I firstly changed the JVM version for the agent only and noticed that there was a substantial improvement so I then proceeded to change the JVM on the server process as well for maximum efficiency. Yes we run the agnet and server on the same physical box which is in itself a single VM on an ESX 4 host. I will investigate stunnel when I can find some time. thanks a lot.