GCPPromd provides Google Compute Engine auto-discovery for Prometheus.

Usage of gcppromd:
  -daemon
    	run the application as a daemon that periodically produces a target file with a json in Prometheus file_sd format. Disables web-mode
  -frequency int
    	(daemon only)  discovery frequency in seconds (default 300)
  -listen string
    	HTTP listen address (default ":8080")
  -outputPath string
    	(daemon only)  A path to the output file with targets (default "/etc/prom_sd/targets.json")
  -projects string
    	(daemon only)  comma-separated projects IDs.
  -projects-auto-discovery
    	(daemon only)  enable auto-discovery of the projects based on which projects can be listed by the provided credentials.
  -projects-excludes string
    	(daemon only) RE2 regex, all projects matching it will not be discovered
  -workers int
    	number of workers to perform the discovery (default 20)

A docker image is available.

docker run messagebird/gcppromd:latest

Outputs a JSON with Prometheus targets in projects (-projects) to a file set by -outputPath.

The http request

GET /v1/gce/instances?projects=<project1,project2,...>

returns an array of prometheus's <static_config>s.

The query parameters are:

• projects accepts a list of coma separated google cloud project names.
• projects-auto-discovery accepts true, 1, TRUE, other values are evaluated to false, add all accessible projects by GCPPromd to the projects list.
• projects-exclude a RE2 regex, all projects matching it will not be discovered.

A "projects auto-discovery" mode can be enabled with -projects-auto-discovery or https://..?projects-auto-discovery=true. In that mode all the accessible projects will be scraped. You can exclude projects using -project-excludes=regex or https://..?project-excludes=regex.

Using the projects auto-discovery add 500ms-1s of overhead to requests/daemon refreshes

The instances on those projects that have the GCE label prometheus (the value doesn't matter) are returned.

Every instance can have one or multiple metadata keys, be careful metadata are not labels, of the form prometheus_ports or prometheus_ports_<service name> mapping to the port number that prometheus should target.

For every GCE metadata like prometheus_ports_* a couple of Prometheus targets and Prometheus labels are emitted.

An instance can also declare the metadata keys prometheus_delegate_address or prometheus_delegate_address_<service name> with prometheus_delegate_ports or prometheus_delegate_ports_<service name>.

These are useful for automatically discovering instances behind a load-balancer.

• __meta_gce_instance_name: the name of the instance
• __meta_gce_metadata_: each metadata item of the instance
• __meta_gce_network: the network URL of the instance
• __meta_gce_private_ip: the private IP address of the instance
• __meta_gce_project: the GCP project in which the instance is running
• __meta_gce_public_ip: the public IP address of the instance, if present
• __meta_gce_subnetwork: the subnetwork URL of the instance
• __meta_gce_tags: comma separated list of instance tags
• __meta_gce_zone: the GCE zone URL in which the instance is running
• __meta_gce_region: the GCE region name in which the instance is running
• __meta_gce_delagate_for_: URLs of the instance delegate.
• __meta_gce_name: the extracted name from the prometheus_port_* GCE label, an empty string if the label is exactly prometheus_port

To authenticate with the google cloud APIs you can use the Application Default Credentials process or set specific credentials using the GOOGLE_APPLICATION_CREDENTIALS environment variable.

Thse credentials need to have the API scope https://www.googleapis.com/auth/compute.readonly.

No errors are ever returned by the API. They are only logged.

Built in support for GCE in prometheus requires you to manually declare all your projects and zones. This can be tedious when you need to map many Google projects across multiple zones.

GCPPromd returns an array of <static_config>, you can automatically write that output to a file and use <file_sd_config>. Prometheus will reload the configuration automatically.

All emitted labels are prefixed with __meta you need to explicitly relabel to pick what you need. For example, assuming that the output of GET /v1/gce/instances is written in /var/local/gcppromd/gcppromd_v1_gce_instances.json:

- job_name: 'gce_auto_discovery'

  file_sd_configs:
    - files:
        - /var/local/gcppromd/gcppromd_v1_gce_instances.json
      refresh_interval: 10m

  relabel_configs:
    - source_labels: ['__meta_gce_name']
      regex: '(.*)'
      target_label: 'gce_name'
      replacement: '$1'
    - source_labels: ['__meta_gce_instance_name']
      regex: '(.*)'
      target_label: 'instance'
      replacement: '$1'
    - source_labels: ['__meta_gce_label_my_label_name']
      regex: '(.*)'
      target_label: 'my_label_name'
      replacement: '$1'