Agent Setup #

The installation steps have to be performed one time, but ensure you keep the agent up to date. It does not provide a public attack surface, but of course new releases usually fix bugs and make the process more stable.

Find the most up to date version of the agent on [our docker registry](https://r.planetary-quantum.com/harbor/projects/7/repositories/collector).

Deploy the agent #

Here’s a full stack template:

version: "3.8"

services:
  collector:
    image: r.planetary-quantum.com/quantum-public/collector:v2.51.2-pq3
    environment:
      CLUSTER: $QUANTUM_ENDPOINT
    networks:
      - metrics
    volumes:
      - agent-wal:/prometheus:rw
      - /var/run/docker.sock:/var/run/docker.sock:ro
    deploy:
      mode: replicated
      resources:
        limits:
          memory: 1G
        reservations:
          memory: 128M
      placement:
        constraints:
          - node.role == manager
      replicas: 1
      restart_policy:
        condition: on-failure

networks:
  metrics:
    attachable: true
    driver: overlay
    name: quantum-custom-metrics

volumes:
  agent-wal:

For multi-manager setups, please update deploy.placement.constraints.

Here’s how to deploy this:

$ export QUANTUM_ENDPOINT=my-cluster
$ export QUANTUM_STACK=${QUANTUM_ENDPOINT}-metrics-collector
$ quantum-cli stack update --create
...

Following the example above, we created a stack called my-cluster-metrics-collector and a network called quantum-custom-metrics. This network is an additional security boundary and allows for private access from the agent process to your tasks and their /metrics endpoint.

Updating the agent #

To update the image, adjust the reference in the stack template and re-deploy:

$ export QUANTUM_ENDPOINT=my-cluster
$ export QUANTUM_STACK=${QUANTUM_ENDPOINT}-metrics-collector
$ quantum-cli stack update
...