At AWS re:Invent yesterday, Matt Williams of Datadog did a demo with The New Stack about how to monitor containers on Amazon Web Services using the Datadog platform.
Customers are asking how to monitor containers on AWS, Williams said. The reason: it is easy to go from one, two to hundreds or thousands of containers. But understanding the role of the container or how it is behaving in an availability zone can be difficult to do.
In the example, Williams showed the increase in containers on a web site and how the load is getting managed. With Datadog, he showed how the user monitors the lifecycle of that increase.
He showed the number of running and stopped containers. Williams also showed how to simulate demand using a Redis benchmarking tool and a distributed net client.
With Datadog, the user can see the number of containers over time and information such as how CloudWatch reports CPU usage across the instances, how the operating system is reporting usage and how Docker is reporting the usage as well.
A different dashboard shows it from a slightly different perspective. For example, it can show the monitoring data by availability zone. Graphs are created by specifying different tags such as CloudWatch or a specific region. Other events might be GitHub commits and the time Docker containers are brought up or destroyed.
A Datadog dashboard can be shared with the world or there are dashboards that can be used internally.
A graph is created by the sum of Docker containers running across all the instances. It can be overlaid by overlaid zones or events as well.