Usage report

The Controller can send, once a day, a short report of what its fleet is made of: how many nodes and clusters, which versions, extensions, HAProxy branches and kernel tracks, how big the nodes are and which features are in use. It is off until an admin turns it on - on the setup screen, or in Settings › Usage report - and changes nothing about how the Controller works.

Why

Janus is built by a small team. The report says where to spend it:

• what to work on next - which versions nodes stay on, which features and platforms are used;
• what could be retired - an extension, a HAProxy branch or a kernel track nobody runs is a cost with no one to pay for it;
• which images to pre-build first - when a release comes out, the image factory builds the update of every schematic in use; the most run go first, and the factory’s mirror (see Images and extensions) holds what was built.

What it holds

Counts, and nothing that names anything: never a node’s or a cluster’s name, an address, a label, a hostname, a certificate, a configuration or an account. The receiver keeps no IP address either - its front end logs nothing for it.

examples/telemetry/report.json

{
  "schema": 1,
  "install_id": "4f9c1e7a2b6d48e0a3c5f17d9e2b6a81",
  "controller": {
    "version": "v2026.10.29-3",
    "arch": "amd64",
    "prerelease": false
  },
  "nodes": {
    "total": 7,
    "unreachable": 1,
    "versions": {
      "v2026.10.29": 2,
      "v2026.10.29-3": 4
    },
    "arches": {
      "amd64": 6
    },
    "haproxy": {
      "3.2": 1,
      "3.4": 5
    },
    "kernel": {
      "stable": 6
    },
    "console": {
      "serial": 6
    },
    "extensions": {
      "bird": 3,
      "letsencrypt": 5,
      "prometheus-node-exporter": 6,
      "qemu-guest-agent": 6
    },
    "schematics": {
      "5a8d6944047f00079ae8cd55b239a47e1561a108f665fcad553bfe6acdad1d03": 3,
      "c2b5e8a41f0d7c39e6a1b4f8d2e7c05a9b3f6d1e8a4c7b2f5e0d9a3c6b1f4e72": 3
    },
    "vcpus": {
      "1": 2,
      "2": 4
    },
    "memory": {
      "1-2GiB": 4,
      "<1GiB": 2
    }
  },
  "clusters": {
    "count": 2,
    "sizes": [
      2,
      3
    ]
  },
  "platforms": {
    "machines": {
      "libvirt": 2,
      "proxmox": 4
    },
    "added": 1
  },
  "features": {
    "sso": true,
    "mfa_required": true,
    "autoscaling": false,
    "cluster_bgp": true,
    "cluster_vrrp": false,
    "log_forwarding": true,
    "lets_encrypt": true,
    "s3_backups": true,
    "fleet": true,
    "consul": false
  }
}

Field	What
schema	the report’s format, 1
install_id	128 random bits naming this Controller’s reports - nothing else: not the Controller’s own ID, which tags its machines. New ID in Settings makes another, and asks the receiver to delete what it holds under the old one
controller	the Controller’s version, architecture, and whether it is a pre-release
nodes.total, nodes.unreachable	how many nodes the Controller manages, and how many didn’t answer within 4 seconds (counted nowhere else)
nodes.versions, arches, haproxy, kernel, console, extensions	how many nodes run each Janus version, architecture, HAProxy branch and kernel track (as their image was built), console, extension
nodes.schematics	how many nodes run each schematic: its ID is the SHA-256 of the choices above - the image factory pre-builds the most run first
nodes.vcpus, nodes.memory	the nodes’ size, by bucket: 1, 2 (2-3), 4 (4-7) or 8+ vCPUs; <1, 1-2, 2-4 or 4+ GiB
clusters	how many clusters, and how many members each has
platforms.machines	the machines the Controller created, by kind of hypervisor; platforms.added, the nodes added to it otherwise
features	on or off, nothing of how: single sign-on, a second factor required, auto-scaling, a cluster managing BGP or VRRP, log forwarding, Let’s Encrypt through the Controller, backups to S3, the fleet PKI, Consul

Exactly what is sent in the Settings card shows the report as it would leave now.

When and where

• Within an hour of turning it on (a moment the ID picks: Controllers don’t all report at once), then every 24 hours; a report that fails is tried again an hour later, then the next day. Send now sends one at once.
• POST https://janus-telemetry.sw-servers.net/v1/report, over HTTPS, through the proxy the Controller’s environment names (HTTPS_PROXY), never following a redirect.
• Each send is in the Controller’s log (telemetry: report sent), each change of the setting in its audit.

Turning it off for good

JANUS_CONTROLLER_TELEMETRY=off (dashboardd -telemetry off) keeps it off whatever the setting says, and the page can’t turn it on - for a Controller set up from code. JANUS_CONTROLLER_TELEMETRY_URL (-telemetry-url) sends it elsewhere.