mirror of https://github.com/okieselbach/AutopilotMonitor.git synced 2026-10-24 04:32:32 +00:00

Files

T

Oliver Kieselbach 2f9a404323 Record session start and end on the server clock (no reader yet)

Session durations come from device times, and a device clock can be off by
hours for a whole session or change in the middle of one. Every upload
carries the device send time next to the server receive time; their
difference is the device clock error. This change records the session's
start, WhiteGlove Part 2 start and end on the server clock next to the
device values. Nothing reads them yet, so no stored value and no display
changes.

- ServerTime: the rule, defined by tests/fixtures/server-time/cases.json and
  run by both the backend (ServerTime.cs) and the web (lib/serverTime.ts).
- Ingest: ServerTime.Tracker, the rule fed one upload at a time, keeps its
  state in the primary-only column ServerTimeState and writes
  StartedAtServer and ResumedAtServer in the counter merge (no extra read
  or write). It follows a session from its first upload only, and only when
  that upload carries agent_started; sessions already under way when this
  build arrives get their values at completion. On every upload prefix of
  every followed case it gives the rule's start and Part 2 start; a retried
  upload with agent_started is not counted as a restart.
- Terminal: the counter reconcile, which every terminal path runs, also
  computes the rule over the same event scan and writes StartedAtServer,
  ResumedAtServer and CompletedAtServer. Fail-soft: a rule failure never
  costs the counter reconcile or the ingest write.
- Re-registration carries the four columns from the fresh CAS read through
  the Replace (the agent's first upload often races its registration).
- SessionSummary gets the three fields; the index mirrors them, the index
  merge drops primary-only columns; manifests and wire types regenerated.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

2026-10-05 22:53:35 +02:00

cases.json

Record session start and end on the server clock (no reader yet)

2026-10-05 22:53:35 +02:00

README.md

Record session start and end on the server clock (no reader yet)

2026-10-05 22:53:35 +02:00

README.md

Server-time rule — shared cases

Device clocks are wrong in a measurable share of sessions: off by whole hours for the entire session (time zone in the hardware clock), stepped mid-session by a program or a sync, or back on a different era after a restart. Durations therefore come from server time, and cases.json is the one definition that every implementation runs:

Backend — ServerTime in src/Backend/AutopilotMonitor.Functions/Helpers/ServerTime.cs (test: ServerTimeParityTests.cs)
Web — lib/serverTime.ts in src/Web/autopilot-monitor-web (test: lib/__tests__/serverTime.test.ts)

A case is one session: its events with the device event time t, the upload's device send time sent (X-Send-Time-Utc, null for agents without it) and the server receive time recv, plus completedAt (the stored device-clock completion, null while running) and isPreProvisioned. Every event of one upload shares sent and recv.

The rule:

Offset of an upload = sent − recv (device clock error minus upload latency). An upload with agent_started opens the next agent run. Per upload, the offset in use is the median of its own and the two previous uploads of the same run; the first two uploads of a run use the median of the run's first three. One delayed upload never moves it.
Start = the earliest event time, as today (anchor-eligible sources, never more than 2 h before the first agent_started), minus the start offset (median of the first three uploads of run 1). Candidates come only from run 1 while its offset stays within 60 s of the start offset. An upload whose own offset is outside that band counts only once the next upload is back inside it — a late upload, not a clock step.
End = the live activity event (0 ≤ sent − t ≤ 60 s, not a history watcher) at completedAt, else the last one before it, minus its offset.
Part 2 start (part2Start) = the first agent_started after the first whiteglove_complete, minus the median of that run's first three uploads; reported whenever both exist, also while running and when Part 2 does not fit the duration.
WhiteGlove (isPreProvisioned) — Part 1 = first whiteglove_complete − start; Part 2 = end − Part 2 start, applied only when the Part 2 start is not after the end.
Event server time (eventServerTimes) = event time minus its upload's offset in use.

While a session runs, the backend keeps the rule's streaming form on the session row (ServerTime.Tracker, column ServerTimeState): it absorbs one upload at a time and writes StartedAtServer and ResumedAtServer. It follows a session only from its first upload, and only when that upload carries agent_started; other sessions get no live values. ServerTimeParityTests checks the tracker against the rule on every upload prefix of every followed case. Once CompletedAt is set, the backend computes the rule once more over the session's stored events and writes StartedAtServer, ResumedAtServer (the Part 2 start) and CompletedAtServer.

expect holds start, startOffset (seconds), end, duration, part1, part2 (seconds), part2Start and the probed event server times; null where the input does not allow a value (no sent, still running). Expectations come from the reference implementation the rule was measured with; every value must be reproduced to the millisecond. A change to the rule goes into cases.json first; the two tests then say which side is behind.

README.md Unescape Escape

Server-time rule — shared cases

README.md