Magnemite
Features help

Automatic updates

New version, cached, rolled out inside your time window, with the guardrails that make that safe.

Settings → Auto-update. When it is on, the hub:

  1. polls both sources every SOURCE_POLL_MINUTES,
  2. caches the newest approved version onto the server,
  3. creates a rollout with your canary and soak settings,
  4. dispatches only inside the time window, if you set one.

What it will not do

  • Never downgrades. A source publishing an older build is ignored.
  • Never runs two rollouts for the same app at once.
  • Does nothing at all if every device is already on the newest version.
  • Only touches approved versions (see below).

Approval

Leaving approve new versions automatically off means a human ticks approve on the Versions page before anything moves. That is a reasonable default while you are still learning to trust it.

With it on, the pipeline is fully hands-off: a release appears upstream and the fleet is on it.

Auto-approve plus no canary means an upstream mistake reaches 200 boxes unattended. Keep a canary of at least a few devices, and a soak long enough that a scanner failure would show up in Rotom before the rest go.

The time window

A window is the difference between 35 GB of egress at 04:00 and 35 GB during your peak scanning hours. Jobs created outside it stay queued; the scheduler dispatches them when it opens.

Polling

SOURCE_POLL_MINUTES defaults to 15 in the compose files. .env.example ships a much lazier 10080 (7 days), because we don't need to update that often.

Both sources are polled on the same tick, and a poll can be forced from the Versions page.

On this page