Monitoring

CloudWatch Alarms

Move to CloudWatch Alarms.

  1. Target tracking may be in Insufficient data state. A target tracking scaling policy does NOT perform scaling when the specified metric has insufficient data. InsufficientData
  2. If the state does not change into In alarm, continue to run cluster_loadtest.sh script.
$ ./cluster_loadtest.sh 

InAlarm

  1. Click the alarm name and you can see Networking > 50000 above the red baseline. AlarmGraph

Amazon ECS

  1. Open Amazon ECS in another browser tab and monitor ECS clusters. Open ECS Instances tab.
  2. You can see the number of ECS instances becomes 4. You can see the newly added instances do not have running tasks yet. ClusterAutoScaled
  3. If a new task provisioned, ECS select its instance with fewer tasks. SpreadTasks

What we have learnt in 5. Monitoring

  1. Select Metrics tab. You can find basic information and let**s find more by clicking View Container Insights. It will redirect you to CloudWatch Container Insight. ECSMetrics
  2. Container Insights shows more detailed information about your ECS clusters. You can find network traffic (bytes/second) metrics too, which is ECS cluster autoscaling trigger. If you put your mouse on the graph, you can see more specific information of time and bytes. NetworkInsights