Container Metrics Exporter

Exports Cloudflare Containers metrics to Datadog using the Cloudflare GraphQL Analytics API. Runs as a Cloudflare Workflow triggered every minute to scrape metrics for all containers in an account.

Setup

Requirements

• Node.js 24+
• npm

Installation

git clone https://github.com/flarelabs-net/cloudflare-container-metrics-datadog-exporter.git
cd cloudflare-container-metrics-datadog-exporter
npm install

Configuration

npm run token-setup

This will prompt for your Cloudflare Account ID, open the browser to create an API token with the required permissions, and collect your Datadog API Key. It automatically creates a .dev.vars file.

Optional: Jurisdiction

To run GraphQL queries from a specific jurisdiction (closer to the data source), set the JURISDICTION variable in wrangler.jsonc:

"vars": {
  "BATCH_SIZE": 5000,
  "RETRY_LIMIT": 3,
  "RETRY_DELAY_SECONDS": 1,
  "JURISDICTION": "eu", // e.g., "eu", "fedramp"
  "DATADOG_TAGS": {}
}

This uses a Durable Object to proxy requests from the specified jurisdiction.

Optional: Custom Datadog Tags

Add custom tags to all metrics by setting the DATADOG_TAGS variable in wrangler.jsonc:

"vars": {
  "BATCH_SIZE": 5000,
  "RETRY_LIMIT": 3,
  "RETRY_DELAY_SECONDS": 1,
  "JURISDICTION": "eu",
  "DATADOG_TAGS": {
    "env": "production",
    "team": "platform",
    "service": "containers"
  }
}

These tags will be added to all health and resource metrics sent to Datadog.

Verify

npx wrangler dev

In another terminal run:

curl "http://localhost:8787/cdn-cgi/handler/scheduled"

You should see an "Ok" response from the curl and logs from the exporter.

Deploy

npx wrangler deploy
npx wrangler secret bulk .dev.vars

Metrics

All metrics are prefixed with cloudflare.containers..

Instance Health (per-application)

Metric	Type	Description
instances.active	gauge	Number of active instances
instances.healthy	gauge	Number of healthy instances
instances.stopped	gauge	Number of stopped instances
instances.failed	gauge	Number of failed instances
instances.scheduling	gauge	Number of scheduling instances
instances.starting	gauge	Number of starting instances
instances.max	gauge	Max instances configured

Tags: account_id, application_id, application_name

Instance Health (global totals)

Metric	Type	Description
instances.total.active	gauge	Total active instances
instances.total.healthy	gauge	Total healthy instances
instances.total.stopped	gauge	Total stopped instances
instances.total.failed	gauge	Total failed instances
instances.total.scheduling	gauge	Total scheduling instances
instances.total.starting	gauge	Total starting instances
instances.total.max	gauge	Total max instances

Tags: account_id

Use instances.total.max - instances.total.healthy to calculate available capacity.

Resource Metrics

Metric	Type	Unit	Description
cpu	gauge	vCPU	CPU load
memory	gauge	MiB	Memory usage
disk	gauge	MB	Disk usage
bandwidth.rx	count	bytes	Bytes received
bandwidth.tx	count	bytes	Bytes transmitted

Tags: account_id, application_id, application_name, version, instance_id, placement_id, stat

• version - The container application version number
• instance_id - The instance identifier (maps to Cloudflare's deploymentId)
• placement_id - The placement identifier (specific realization of an instance, useful for tracking restarts/churn)
• stat - The aggregation type: p50, p90, p99, max (bandwidth metrics don't have a stat tag)

Datadog Dashboard

A pre-built dashboard is included in datadog-dashboard.json. To import it:

1. In Datadog, go to Dashboards → New Dashboard → New Dashboard
2. Click the cog icon (⚙️) in the top right
3. Select Import dashboard JSON
4. Paste the contents of datadog-dashboard.json

See Datadog's documentation for more details.

Workflow Behavior

The exporter runs as a Cloudflare Workflow triggered every minute via cron. Each workflow step uses configurable retry settings:

• Retries: Configurable via RETRY_LIMIT (default: 3 attempts)
• Delay: Configurable via RETRY_DELAY_SECONDS (default: 1 second initial delay)
• Backoff: Exponential (e.g., 1s, 2s, 4s)

Steps will automatically retry on transient failures (API errors, network issues).

Configuration Options

Variable	Type	Default	Description
BATCH_SIZE	number	5000	Maximum metrics per Datadog API request
RETRY_LIMIT	number	3	Number of retry attempts for failed workflow steps
RETRY_DELAY_SECONDS	number	1	Initial delay in seconds before retry (exponential backoff)
JURISDICTION	string	""	Durable Object jurisdiction for GraphQL queries (e.g., "eu", "fedramp")
DATADOG_TAGS	object	{}	Custom tags to add to all metrics