Skip to main content
Version: master

Technical Overview

How Pantavisor works and why it is built this way, ordered to be read as a book. For the exact keys, fields and endpoints behind any of it, follow the Reference links at the foot of each page, or start from the Reference index.

Foundations​

  1. Architecture — the two jobs Pantavisor does, and the state machine it does them in
  2. Revisions — the unit of state: what a revision is and how state JSON makes it reproducible
  3. BSP — kernel, modules, firmware and bootloader, versioned alongside application containers
  4. Containers — runtime, groups, roles, status, auto-recovery and lifecycle control

State and storage​

  1. Updates — the atomic update flow, its progress states, and the failure paths
  2. Storage — on-disk layout, object store, logs, metadata and integrity
  3. Disks — disk types, single and dual-mode partitioning, dm-crypt, boot sequence

Control​

  1. Remote Control — the Pantacor Hub client and other remote controllers
  2. Local Control — the pv-ctrl socket, Pantabox and pvcontrol

Runtime services​

  1. IPAM — container IP address management: pools, allocation, network namespaces
  2. Inter-Container Communication — the xconnect service mesh
  3. Hooks — lifecycle hooks, and how system-start gates a revision
  4. Watchdog — kernel watchdog integration and the four ping modes
  5. Power and Wakelocks — blocking and scheduling suspend without missing updates

Configuration and modes​

  1. Configuration Levels — the precedence hierarchy across factory, device and container scopes
  2. Init Mode — embedded, standalone and appengine, and when to use each