Amazon EC2 Instance Monitoring?

Question

I am in need of a fairly short/simple script to monitor my EC2 instances for Memory and CPU (for now). After using Get-EC2Instance -Region , it lists all of the instances. from here where can i go?

Answer 1

Cloudwatch is the monitoring tool for AWS instances. While it can support custom metrics, by default it only measures what the hypervisor can see for your instance.

CPU utilization is supported by default, this is often a more accurate way to see your true CPU utilization since the value comes from the hypervisor.

Memory utilization however is not. This depends largely on your OS and is not visible to the hypervisor. However, you can set up a script that will report this metric to Cloudwatch. Some scripts to help you do this are here: http://docs.aws.amazon.com/AmazonCloudWatch/latest/DeveloperGuide/mon-scripts-perl.html

Answer 2

There are a few possibilities for monitoring EC2 instances.

• Nagios - http://www.nagios.com/solutions/aws-monitoring
• StackDriver - http://www.stackdriver.com/
• CopperEgg - http://copperegg.com/aws/

But my favorite is Datadog - http://www.datadoghq.com/ - (not just because I work here, but its important to disclose I do work for Datadog.) 5 hosts or less is free and I bet you can be up and running in less than 5 minutes.

Answer 3

Depends what your requirements are for service availability of the monitoring solution itself, as well as how you want to be alerted about host/service notifications.

Nagios, Icinga etc... will allow you to customise an extremely large number of parameters that can be passed to your EC2 hosts, specifying exactly what you want to monitor or check up on. You can run any of the default (or custom) scripts which then feed data back to a central system, then handle those notifications however you want (ie send an email, SMS, execute an arbitrary script). Downside of this approach is that you need to self-manage your backend for all of the aggregated monitoring data.

The CloudWatch approach means your instances can push metric data into AWS, then define custom policies around thresholds. For example, 90% CPU usage for more than 5 minutes on an instance or ASG, which might then push a message out to your email via SNS (Simple Notification Service). This method reduces the amount of backend components to manage/maintain, but lacks the extreme customisation abilities of self-hosted monitoring platforms.