0

Is there a way to monitor ec2 instances without using cloudwatch? I ask this since ec2 instances are basically VPS's, and using output from commands like top, vmstat, htop in scripts may not give the clear picture as the CPU cycles are shared between other instances as well.

What should one keep in mind while monitoring CPU usage on a VPS?, should one have alerts based on top load or % of CPU used by user processes coupled with other factors like processes waiting on disk io, hardware interrupts?

sektor
  • 13
  • 3

1 Answers1

1

Monitor it like any other server (Nagios, opennms etc). The output of top etc. will be specific to your container, rather than the hypervisor.

As for the second part of your question; that is personal preference. I would not personally set up alerting based off CPU load, as it does not definitively indicate a problem.

Peter
  • 1,450
  • 2
  • 15
  • 26
  • From what I've read in these posts, people disagree about using top as a measurement tool in VPSes. http://serverfault.com/questions/420122/why-does-top-report-a-different-cpu-usage-than-cloudwatch http://serverfault.com/questions/400533/why-does-ec2-monitoring-show-100-cpu-and-top-only-20?rq=1 – sektor Aug 19 '14 at 15:01