monitor.factory

Shared monitor-collector factory.

Used by both otto monitor (live dashboard) and otto test --monitor (session-scoped collection during a test run). Centralising this here keeps both call sites consistent — same parser-registry lookup, same per-host log silencing, same target construction.

otto.monitor.factory.build_monitor_collector(hosts: Sequence[RemoteHost], db: MetricDB | None = None, tunnel_source: Callable[[], Awaitable[list[TunnelRecord]]] | None = None) → MetricCollector

Build a MetricCollector over hosts.

Creates one MonitorTarget per host, silences host logging (collection is chatty), and chooses each host’s collection mode:

  • a host with an snmp block is polled over SNMP — its SnmpOptions becomes a live SnmpClient (address defaulting to the host’s own ip) plus the OID list to GET;

  • otherwise it is polled by running shell commands, with its parser set resolved via get_host_parsers() so per-host customisations registered by init modules are honoured.

Parameters:

  • hosts – Hosts to sample on each tick.

  • db – Optional session-bound MetricDB for persistence (unopened); None means in-memory. The frame (session identity) is the caller’s job — this factory stays session-blind, same as the collector it builds.

  • tunnel_source – Optional full-lab tunnel discovery callable for the collector’s tunnel loop (spec 2026-07-16). The CLI composes this over the WHOLE lab — tunnels may traverse hosts that metric collection was never pointed at.