Overview
Containers are supported in Pro and Enterprise plans. Depending on your plan, you can monitor 5 or 10 containers free for each host license. The container count is averaged across your entire infrastructure.
Additional containers are billed at an additional cost per container per hour. In addition, you can purchase prepaid containers. Contact Sales or your Customer Success Manager to discuss containers for your account.
Kubernetes
Kubernetes creates pause containers (requires Agent v5.8+) to acquire the respective pod’s IP address and set up the network namespace for all other containers that join that pod. Datadog excludes all pause containers from your quota and does not charge for them (requires Agent v7.20+ for AWS EKS pause container exclusion).
Fargate
Fargate is charged based on the concurrent number of monitored tasks in ECS Fargate and the concurrent number of monitored pods for EKS Fargate. For pricing, see the Infrastructure section on the pricing page.
GKE Autopilot
Billing of GKE Autopilot environments is the same as that of GKE Standard.
Frequently asked questions
How does Datadog measure hourly on-demand usage?
Containers are metered in five minute increments. The difference is calculated between the number of containers observed and the allotment for the account. The allotment for the account is the total of the included containers (5/host for Pro and 10/host for Enterprise by default) and any contracted container commitment.
First, the five-minute usage increments are averaged over the hour by dividing total usage by twelve. Secondly, the allotment is deducted from each hour. Finally, the total on-demand usage is the sum of the on-demand usage for each hour. This method normalizes the on-demand usage for short term spikes and variations between hosts.
What if a user runs a particularly high number of containers for a short period of time?
Using the hourly on-demand usage calculation above, if there are 1200 on-demand containers during a single five-minute interval (1/12 of an hour), these containers are metered as 100 container hours (1200/12).
What if containers are unevenly distributed across hosts?
As long as the number of total running containers does not exceed the total allotment for the account, these are included in the account’s infrastructure plan.
When using autoscaling groups, there are have peak and off-peak hours. How does this affect the on-demand container usage?
The on-demand calculations are performed using the allotment that is based on the number of infrastructure hosts. Since the autoscaling groups increase the number of hosts during peak hours, the total container allotment also increases during those periods.
Are Datadog Agent containers counted against the allocation?
No, the Datadog Agent containers are not counted against your 5 (Pro tier) or 10 (Enterprise tier) allocation.
Are containers belonging to pods in constant CrashLoopBackoff counted?
If a container in the pod is running for over ten seconds, it is counted against the allotment during the metering interval. Since the pod is unhealthy, the data collected in these situations is valuable for troubleshooting. Also, the data is tagged by the pod name, so as underlying containers attempt to re-run (each time as a new container_id) all of the context (such as relevant events and log entries) is kept together in the pod view.
Troubleshooting
For technical questions, contact Datadog support.
For billing questions, contact your Customer Success Manager.
