Resolution
I want to monitor OS metrics and performance counters of my Amazon
Elastic Compute Cloud (EC2) Linux instance. How can I do this using
Amazon CloudWatch?
Short Description
You can create a custom CloudWatch metric for your EC2 Linux instance
statistics by creating a script through the AWS Command Line Interface
(AWS CLI). Then, you can monitor that metric by pushing it to
CloudWatch.
Resolution
Before proceeding, be sure that you install and configure the AWS CLI for use with the instance that you want to monitor.
Create a custom CloudWatch metric
To create your custom metric:
Log in to the instance through the AWS CLI.
Copy the following bash script, and then save it to your instance (for example, mem.sh).
This example script shows the values that you can publish in CloudWatch. In this example, the put-metric-data API call is used to push the following values to CloudWatch:
- Percentage of used memory (USEDMEMORY)
- Number of total connections (TCP_CONN)
- Number of TCP connections on port 80 (TCP_CONN_PORT_80)
- Number of users currently logged in (USERS)
- Percentage of I/O wait time (IO_WAIT)
========Sample script======
#!/bin/bash
USEDMEMORY=$(free -m | awk 'NR==2{printf "%.2f\t", $3*100/$2 }')
TCP_CONN=$(netstat -an | wc -l)
TCP_CONN_PORT_80=$(netstat -an | grep 80 | wc -l)
USERS=$(uptime |awk '{ print $6 }')
IO_WAIT=$(iostat | awk 'NR==4 {print $5}')
aws cloudwatch put-metric-data --metric-name memory-usage --dimensions Instance=i-0c51f9f1213e63159 --namespace "Custom" --value $USEDMEMORY
aws cloudwatch put-metric-data --metric-name Tcp_connections --dimensions Instance=i-0c51f9f1213e63159 --namespace "Custom" --value $TCP_CONN
aws cloudwatch put-metric-data --metric-name TCP_connection_on_port_80 --dimensions Instance=i-0c51f9f1213e63159 --namespace "Custom" --value $TCP_CONN_PORT_80
aws cloudwatch put-metric-data --metric-name No_of_users --dimensions Instance=i-0c51f9f1213e63159 --namespace "Custom" --value $USERS
aws cloudwatch put-metric-data --metric-name IO_WAIT --dimensions Instance=i-0c51f9f1213e63159 --namespace "Custom" --value $IO_WAIT
===============================================
- After creating the bash script, give execute permissions to the file.
$ chmod +x mem.sh
- Run the bash script to check that it works.
Push your metric to CloudWatch
- Create a cron job:
$ crontab -e
- Add this line to execute your script every minute:
*/1 * * * * /home/ec2-user/mem.sh
- Save and exit.
When the crontab is saved, crontab: installing new crontab appears.
Monitor your EC2 instance
Find your custom metric in the CloudWatch console:
Open the CloudWatch console.
Choose Metrics.
Choose the All Metrics tab.
Choose Custom.
Choose the dimension Instance.
Select your custom metric by its InstanceId and Metric Name.
View the graph of your metric.
Other uses
You can use this example to build your own logic to process multiple dimensions, and then push that metric data to CloudWatch.
For example, suppose that you benchmark your application. Then, you
discover that when the I/O wait time and percentage memory usage reach a
certain threshold, the system stops functioning properly. To address
this problem, you can monitor both values simultaneously in a script.
Store the logical AND of the values in a third variable that you push to
CloudWatch.
c=0
if [[ $IO_WAIT > 70 && $USEDMEMORY > 80 ]]
then
c=1
fi
aws cloudwatch put-metric-data --metric-name danger --dimensions Instance=i-0c51f9f1213e63159 --namespace "Custom" --value $c
For normal conditions, this variable is 0 (zero). For situations when
both conditions are met, the value is set to 1 (one). You can then
build custom alarms around these parameters to warn of problematic
situations for your system.