Amazon ECS Cluster
Does ZopNight manage Amazon ECS Cluster?
An ECS cluster is free as an object. The bill is the EC2 container instances and Fargate tasks its services keep running. ZopNight discovers clusters every 6 hours and parks non-production ones by scaling every service in the cluster to a desired count of 0 on schedule, restoring counts at start.
Rules that fire on Amazon ECS Cluster
At a glance
| Field | Value |
|---|---|
| Scheduling notes | every service in the cluster is scaled to a desired count of zero on stop and restored on start. |
An Amazon Elastic Container Service (ECS) cluster is the logical grouping for containerized services and tasks. The cluster itself is free, but the EC2 or Fargate capacity its services keep warm bills continuously, so non-production clusters are strong scheduling candidates.
A free wrapper around paid capacity
The cluster object costs nothing and has no meter of its own. What bills is the capacity registered to it: container instances on the EC2 launch type bill per instance-hour whether tasks land on them or not, and Fargate tasks bill per vCPU-hour and GB-hour while they run. The cluster is where that capacity is organized, which makes it the natural unit to park when an entire environment sleeps.
Parking an environment in one move
At the scheduled stop, ZopNight scales every service in the cluster to a desired count of zero, saving each service’s previous count; at start, every saved count is restored. This is the cluster-level equivalent of stopping each service individually, and it suits dev and staging clusters that map one-to-one to an environment. On Fargate the parked cost drops to zero; on the EC2 launch type the container instances need their own schedule to finish the job, since they bill independently of task placement.
Two cluster-shaped waste patterns
The empty cluster: container instances registered and billing with zero services and zero running tasks, the usual leftover of a migration that moved workloads to a new cluster and forgot the old one’s capacity. And the always-on environment: a cluster whose services do their work in roughly 45 hours a week while its capacity is provisioned for all 168.
Discovery and attribution
Clusters are found every 6 hours via Resource Explorer 2 and the ECS API, with cost attributed through the underlying compute. Recommendations cover idle capacity and rightsizing across the cluster’s services.
The cluster dashboard
ECS console, then Clusters. The cluster page shows registered container instances, service count, and running task count together. A nonzero instance count sitting over a zero task count is the empty-cluster signature.