const SUSTAINED_FAILURE_THRESHOLD_SECS: i64 = 120;Expand description
Sustained-failure window cap. RFC-0005 §6 — the agent transitions Soaking → Failed when a probe has been failing continuously past this threshold.
TODO(v0.2.1): wire from services.nixfleet-agent.healthChecks via
the NixOS module → agent CLI arg → runtime config struct, per
RFC-0005 §6 + §9.1 (“each agent reads it from
services.nixfleet-agent.healthChecks”). Bigger than v0.2 scope
because it touches the NixOS module surface; v0.2 ships with the
hardcoded floor.
Tracking issue: open on abstracts33d/nixfleet (Forgejo origin
or GitHub upstream, operator’s call) — title “Wire
SUSTAINED_FAILURE_THRESHOLD_SECS from NixOS module config (v0.2.1)”.
The hardcoded 120s is twice RFC-0005 §6’s documented default
(60s), so under-shooting safely: real probe-failure detection
still fires, just 60s later than a tuned deployment would. Safe
for v0.2 demo + lab work; not appropriate for production fleets
with tight SLOs. NixOS-module wire-through to make this
operator-tunable is tracked in v0.2.1-followups.md.