I tried to get CloudWatch data with Python

Trigger

Every day I check the status of the server on the CloudWatch dashboard. When you come to work, first check the dashboard in order ... It's annoying! : persevere: I want to get dashboard information at once! Alright, let's write in Python!

Get CloudWatch information with Python

It would be nice to use Boto3's get_metric_statistics (). As per Official Documentation, first prepare to load CloudWatch with Boto3.

import boto3
client = boto3.client('cloudwatch')

It seems that get_metric_statistics () is used as follows.

response = client.get_metric_statistics(
           Namespace = 'string',
           MetricName = 'string',
           Dimensions = [
             {
               'Name': 'string',
               'Value': 'string'
             },
           ],
           StartTime = datetime(2020, 2, 11),
           EndTime = datetime(2020, 2, 11),
           Period = 123,
           Statistics = [
             'SampleCount'|'Average'|'Sum'|'Minimum'|'Maximum',
           ])

Namespace ʻAWS / EC2 or ʻAWS / ElastiCache or ʻAWS / RDS. It's written at the top of the information that appears when you hover your mouse cursor over the detailed information on the CloudWatch Graphed Metrics` tab.

MetricName CPUUtilization, MemoryUtilization, DiskSpaceAvailable, etc. It's written above the separator line, the second from the top of the information that appears when you hover your mouse cursor over the details on the CloudWatch Graphed Metrics tab.

Dimensions ʻInstanceId, CacheClusterId, DBInstanceIdentifier, etc. It's written below the information separator that appears when you hover your mouse cursor over the detailed information on the CloudWatch Graphed Metricstab. (All dimensions information below the dividing line) As in the example above, Dimensions is written in the following form.Dimensions=[{'Name': 'string', 'Value': 'string'}] So specifically Dimensions = [{'Name':'InstanceId','Value':'i-xxxxxxxxxxx'}] Dimensions=[{'Name': 'Role', 'Value': 'WRITER'}, {'Name': 'DBClusterIdentifier', 'Value': 'xxxxxxxxxxx'}]` It will be shaped like.

Period Write the period in seconds. So 1 minute → 60 5 minutes → 300 24 hours → 86400

Statistics Write statistics.

response

The response of get_metric_statistics () is as follows.

{'Label': 'CPUUtilization', 'Datapoints': [{'Timestamp': datetime.datetime(2020, 2, 10, 19, 8, tzinfo=tzutc()), 'Maximum': 6.66666666666667, 'Unit': 'Percent'}], 'ResponseMetadata': {'RequestId': 'xxxxxxxxxxx', 'HTTPStatusCode': 200, ...(abridgement)

So The value is response ['Datapoints'] [0] [value specified in Statistics (Maximum in the above response)] The unit of the value is response ['Datapoints'] [0] ['Unit'] It seems that you can make it feel good if you use around.

What I actually wrote (part)

Since I was visually checking with CloudWatch every day, in this script I will get the past 24 hours from the script execution date and time. Once you have a script, you may want to run it regularly with cron.

import boto3
from datetime import datetime, timedelta

client = boto3.client('cloudwatch')

def get_metric_statistics(name_space, metric_name, dimensions_values, statistic):
    #Get CloudWatch Information
    response = client.get_metric_statistics(
               #For CPU usage`AWS/EC2`Enter
               Namespace = name_space,
               #For CPU usage`CPUUtilization`Enter
               MetricName = metric_name,
               # `[{'Name': 'InstanceId', 'Value': instance_id}]`Enter
               Dimensions = dimensions_values,
               #Start date and time`Script execution date and time-1 day`Specified by
               StartTime = datetime.now() + timedelta(days = -1),
               #End date and time`Script execution date and time`Specified by
               EndTime = datetime.now(),
               #Specify 24 hours
               Period = 86400,
               # `Maximum`Enter
               Statistics = [statistic]
               )

    #Output statement creation
    response_text = name_space + ' ' + metric_name + statistic + ': ' + str(response['Datapoints'][0][statistic]) + ' ' + response['Datapoints'][0]['Unit']
    print(response_text)

#Output target metric
instance_id = 'i-xxxxxxxxxxx'
#CPU usage
get_metric_statistics('AWS/EC2', 'CPUUtilization', [{'Name': 'InstanceId', 'Value': instance_id}], 'Maximum')
#Memory usage
get_metric_statistics('System/Linux', 'MemoryUtilization', [{'Name': 'InstanceId', 'Value': instance_id}], 'Maximum')

Output result

AWS/EC2 CPUUtilizationMaximum: 6.66666666666667 Percent
System/Linux MemoryUtilizationMaximum: 18.1909615159559 Percent

Summary

I tried to get CloudWatch information from a Python script. This time, I used a pattern with only one Datapoints, but depending on the specified period, there will be multiple outputs. In that case, please aim at the necessary data with a good feeling by loop processing.

We're hiring! We are developing an AI chatbot. If you are interested, please feel free to contact us from the Wantedly page!

Reference article

Boto3 Docs CloudWatch