Monitor Docker containers

Collect per-container CPU, memory, network, and state metrics from the Docker daemon and write them to InfluxDB 3.

Configuration

[[inputs.docker]]
  ## Connect to the local Docker daemon socket.
  endpoint = "unix:///var/run/docker.sock"

  ## Collect from running containers only.
  container_state_include = ["running"]

  ## Report total container CPU instead of one metric per core.
  perdevice_include = []
  total_include = ["cpu", "blkio", "network"]

  ## Keep only the compose service label as a tag.
  docker_label_include = ["com.docker.compose.service"]

[[outputs.influxdb_v3]]
  urls = ["http://localhost:8181"]
  token = "
AUTH_TOKEN
"
  database = "
DATABASE_NAME
"

Replace the following:

• AUTH_TOKEN: your InfluxDB authorization token
• DATABASE_NAME: the database to write to

Where to run Telegraf

Telegraf can monitor Docker from the host or from a container of its own. Either way, the Telegraf process must be able to reach the Docker daemon socket that endpoint points to.

On the host: add the telegraf user to the docker group so it can read the socket, then restart Telegraf:

sudo usermod -aG docker telegraf

In a container: use the official telegraf image and bind-mount the Docker socket and your configuration file:

docker run --detach --name telegraf \
  --volume /var/run/docker.sock:/var/run/docker.sock \
  --volume $PWD/telegraf.conf:/etc/telegraf/telegraf.conf:ro \
  telegraf

Mounting the Docker socket grants the container control over the Docker daemon. Only do this with images you trust.

How it works

• endpoint connects to the Docker daemon through its Unix socket. Use tcp://<host>:<port> for a remote daemon, with the TLS options as needed.
• container_state_include skips created, exited, and paused containers, so you only pay for containers doing work.
• perdevice_include and total_include control metric granularity. This example reports accumulated totals per container instead of one metric per CPU core, block device, and interface, which keeps series counts low.
• docker_label_include limits which container labels become tags. Container labels are often high-cardinality. Allowlist only the labels you query by.
• The plugin also produces docker and docker_data summary metrics for the engine itself, such as container counts and storage usage.

Example output

docker,engine_host=host1,server_version=27.0.3 n_containers=12i,n_containers_running=8i,n_images=41i 1709572230000000000
docker_container_cpu,com.docker.compose.service=api,container_image=example/api,container_name=api-1,container_version=1.4.2,cpu=cpu-total usage_percent=12.7 1709572230000000000
docker_container_mem,com.docker.compose.service=api,container_image=example/api,container_name=api-1,container_version=1.4.2 usage_percent=23.1,usage=498073600i,limit=2147483648i 1709572230000000000
docker_container_net,com.docker.compose.service=api,container_image=example/api,container_name=api-1,container_version=1.4.2,network=total rx_bytes=87234981i,tx_bytes=12873411i 1709572230000000000

Extend this example

• Scope collection with container_name_include or metric filters when you only care about specific services.
• Add inputs.docker_log to collect container logs through the same daemon connection.