AWS Metrics Retention

• Cloudwatch retains metrics data
  • Data points with period of less than 60 seconds (frequency) are available for 3 hours. These are called as high resolution metrics
  • Data points with period of 60 seconds are available for 15 days
  • Data points with period over 300 seconds are available for 63 days
  • Data points with period over 3600 seconds are available for 455 days

AWS Cloudwatch Terms

• Dimensions: A dimension is a name/value pair that is part of identity of metric. Each metric can have upto 10 dimensions
• General Dimensions added are
  • Environment = (Dev/QA/UAT/Prod)
  • Project = (name of project)
  • Version = (Version of the project)
• Statistics:
  • Minimum
  • Maximum
  • Sum
  • Average
  • Sample Count
  • pNN.NN : percentile (p96.54)
• Units: Each statistic has a unit of measure . Example units Bytes, Seconds, Percent etc
• Period: frequency of data collection. Generally defined by number of seconds
• Alarm: We use alarm to initiate some action automatically. Alarms can send notifications to SNS or Auto Scaling policy. Alarm can 3 values
  • Insufficient: Alarm doesnot sufficient data to decide
  • OK: Alarm is not met
  • Alarm: Your condition of alarm is met.

Lets play with Alarms

• Scenario 1: Lets take a metric called as CPU Utilization (Percentage CPU) and when Percentage CPU is less than 40 lets shutdown the system.

• Solution:

• Scenario 2: Lets create an alarm to restart ec2 instance when Percentage CPU is grater than 90% for 5 mins

  • Lets login into ec2 instance and try to install a tool called stress

  sudo amazon-linux-extras install epel
  sudo yum install stress -y
  stress --cpu 8 -v --timeout 100m
  

  • Now watch for CPU Utilization
  • Now create an alarm for above condition

• Scenario 3: Create a cloudwatch alarm with no action to (Exercise)

  • Plot if the DB Connections are less than or equal to 10 in last 1 hour
  • Plot if the free space is less than 100 MB

• Scenario 4: Send an email when cpu utilzation is grater than 80 % for 5 minutes

• Scenario 5: Send an email for underused ec2 instance (when cpu utilization is less than 30)

• Scenario 6: Send an email for under used RDS instance (DB Connections equal to zero in last 1 hour)