Bridges Docker Swarm services to Prometheus via file-based service discovery
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


TL;DR: use -e METRICS_ENDPOINT=/metrics (or METRICS_ENDPOINT=:8080/metrics if using port != 80) for your Docker Swarm service, and Prometheus will automatically discover your service and scrape metrics from it.

Integrates Docker (Swarm) with Prometheus via the file config option (Prometheus currently doesn't have support for Swarm mode).

This image contains the whole bundle, Prometheus + docker-prometheus-bridge + configuration for truly automated service discovery.

Because we're already integrating with Docker, we also export containers' metrics and while we're at it, host's metrics (see /hostproc mount).

+--------------------------+           +----------------------+
|                          | writes    |                      |
| docker-prometheus-bridge +-----------> Configuration file   |
|                          |           | (Scrapeable targets) |
+-------^------------------+           |                      |
        |                              +-----------+----------+
        | queries                                  |
        |                                          |
        |                                          | reads
+-------+------+                                   |
|              |                              +----v-------+
| Docker API   |                              |            |
| (Swarm mode) |                              | Prometheus |
|              |                              |            |
+--------------+                              +------------+

Only advertises services that have environment variable defined: METRICS_ENDPOINT=/metrics (whitelisting).

Uses service name as job name in Prometheus.

NOTE: current limitation is that you have to deploy Prometheus on the same node as a Swarm manager! (use constraint with $ docker service create if/when you have a multi-node cluster)

Howto run

Basically, you need to run this as a Swarm service (for Prometheus to have access to services running in Swarm network):

$ docker service create --network YOUR_NETWORK --name prometheus -p 9090:9090 \
	-e METRICS_ENDPOINT=:8081/metrics \
	--mount type=bind,src=/var/run/docker.sock,dst=/var/run/docker.sock \
	--mount type=bind,src=/proc,dst=/hostproc \
	--mount type=volume,src=prometheus-dev,dst=/prometheus \

Example config file to be fed to Prometheus

This is the file format that docker-prometheus-bridge generated for Prometheus:

		"targets": [
		"labels": {
			"job": "html2pdf"


$ docker run --rm -it -v "$(pwd):/app" golang:1.8.0 bash
$ cd /app
$ go get -d ./...
$ make

# Ctrl + d

You now have ./app statically compiled.

Running $ docker build would bake that into a Docker image.

Alternatives & other reading


  • I'm not happy with this "bridge" binary running inside the Prometheus container - we could impersonate Consul catalog server and integrate via that option over HTTP.
  • Currently the metrics work ins a single-node setting only. We could fix that with a leader + agent -based architecture, where Prometheus only contacts the leader which reports all agents' (runs on each node) IPs as Prom targets, so in result Prometheus could scrape the entire cluster's metrics.