NUC: the stack doesn't start at boot — it waits for Jeff to log in (9.5h outage), and it was never the sleep settings

Aperta
#331 9 commenti 0 reazioni 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
18/100
Tipo di issue
Bug
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Stack tecnologico
docker, powershell

Direzione di ricerca

The payload names no repository files. The boot-time problem lives on the Windows host: the HKCU Run entry for Docker Desktop, the com.docker.service startup type, and the absent scheduled tasks. Start by confirming those states with the commands in the issue, then wait for the owner's decision between auto-login, boot-time startup, or alerting. Done means a real reboot brings containers up within minutes, verified by container StartedAt and boot_dependency_wait.log.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

bug infrastructure priority:high status:blocked triage:parked

Summary

The eight-day "PPS is broken every morning" story had two independent causes stacked, and the one we were all chasing — power/sleep settings — was never one of them.

What the machine actually says

1. Sleep was never the problem. On the active (Balanced) scheme, AC:

Sleep after      (STANDBYIDLE)  : 0x00000000   (never)
Hibernate after  (HIBERNATEIDLE): 0x00000000   (never)

Already disabled. The standing "go fix the Windows power/sleep settings" to-do has been carried through three incidents and would not have helped.

2. Two of the three incidents were Windows Update reboots, not crashes — System log, Event 1074:

When (local) Process Reason
2026-09-14 23:00 TrustedInstaller.exe Operating System: Upgrade (Planned)
2026-09-08 22:31 TrustedInstaller.exe Operating System: Upgrade (Planned)
2026-08-11 22:31 TrustedInstaller.exe Operating System: Upgrade (Planned)
2026-07-14 23:31 TrustedInstaller.exe Operating System: Upgrade (Planned)

Patch Tuesday, monthly, plus a second one this September. Normal and largely not worth fighting — security updates should land.

Only 2026-09-08 09:02 was genuinely unexplained: kernel-power 41 + event 6008 "previous system shutdown was unexpected." One occurrence, no retained evidence. Tracked separately, not a pattern yet.

3. The actual finding: nothing starts at boot.

  • Windows booted 2026-09-14 23:00:13 local.
  • First containers started 2026-09-15 15:29 UTC (08:29 local) — everything else at 15:44 UTC.
  • That is a ~9.5 hour gap, and it lines up with Jeff signing in.

Structurally confirmed, not just inferred from the timing:

com.docker.service                        Stopped / Manual
HKCU\...\CurrentVersion\Run "Docker Desktop"   PRESENT   <- fires at user LOGIN
HKLM\...\CurrentVersion\Run "Docker Desktop"   absent
Scheduled tasks starting docker/wsl at boot    none

So after any reboot, Haven, PPS, ntfy, the daemons and the Observatory are all down until a human physically signs in. Nobody spotted it because by the time Jeff looks, he has just logged in.

Why this changes the picture

#330 (cold-boot dependency race) is correct and necessary — it makes the start clean. But it does not reach this: the outage was never mostly about recovering badly, it was about not starting at all for nine hours. The two fixes compose — #330 makes the start correct, this makes the start happen.

Options (Jeff's call — this is a decision about physical-access posture on his own machine, so nothing here has been changed)

A. Enable auto-login. Reboot → Windows signs in → HKCU Run fires → Docker Desktop starts → containers come up. Turns 9.5h into ~2min.
Tradeoff: anyone with physical access to the NUC gets a logged-in desktop. It is a home machine on his own network, so this is probably fine, but it is his risk to accept.

# shows the auto-login UI (checkbox: "Users must enter a user name and password")
control userpasswords2

B. Leave login required; start the stack at boot instead. Set com.docker.service to Automatic and add a boot-time scheduled task that starts the WSL distro. Keeps the lock screen honest.
Tradeoff: Docker Desktop's engine is session-oriented; this path is fiddlier and wants testing across a real reboot before being trusted.

Set-Service -Name com.docker.service -StartupType Automatic

C. Accept it, and make the blackout visible. Change nothing; add an alert so a multi-hour stack outage is noticed rather than discovered.
Tradeoff: the outage stays; we just stop being surprised by it.

Recommendation: A, with C regardless — the blackout should be visible whatever we choose.

Verification after whichever is chosen

Do not trust it until a real reboot proves it. After the next restart:

powershell.exe -NoProfile -c "(Get-CimInstance Win32_OperatingSystem).LastBootUpTime"
docker inspect -f '{{.Name}} {{.State.StartedAt}}' $(docker ps -aq)
tail ~/.claude/data/boot_dependency_wait.log

Container StartedAt should be minutes after boot, not hours.

Related: #330 (cold-boot race, fixed), #329 (urgency organ — this is the consequence it surfaced first).

Lingua principale
Python
Stelle
2
Fork
0
Metriche di merge delle PR
Nessuna PR unita negli ultimi 30g

Preparare l'ambiente

Questo progetto non fornisce container di sviluppo, Dockerfile né guida per i contributori, quindi l'ambiente è a tuo carico: parti dal suo README e consulta la nostra guida al primo contributo per i passaggi generali.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di JDHayesBC/Awareness

Tutte le issue di JDHayesBC/Awareness

Issue simili

Altre issue su Python

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.