diagnose
/diagnose
Diagnozuje awarie i dziwne zachowanie serwera domowego: usługi, kontenery w pętli restartów, pełny dysk i nieudane aktualizacje.
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 05.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Diagnozuje awarie i dziwne zachowanie serwera domowego: usługi, kontenery w pętli restartów, pełny dysk i nieudane aktualizacje.
Oryginalny opis (EN)
Use when something on the homeserver is broken, erroring, slow or behaving unexpectedly, such as a service down, a container crash-looping, an app unreachable, a full disk, a slow server or a failed update.
Jak działa
- Common causes
Podgląd SKILL.md
---
name: diagnose
description: Use when something on the homeserver is broken, erroring, slow or behaving unexpectedly, such as a service down, a container crash-looping, an app unreachable, a full disk, a slow server or a failed update.
---
# Diagnose
Read `../../shared/conventions.md` (relative to this skill's directory) first and load the server's context. Check the recent `changelog.md` entries first, because most breakage follows a change.
1. **Pin the symptom.** Establish what exactly fails, since when, and for whom: one device or everyone, LAN or tailnet. Reproduce it yourself where you can, for example with `curl` or `docker compose ps`.
2. **Scope it.** Decide whether the problem is host-wide or limited to one service. Run the host checks in one batch: `df -h; df -i; free -h; uptime; systemctl --failed; docker ps -a --format '{{.Names}}\t{{.Status}}'; journalctl -p err -b --no-pager | tail -50; sudo -n dmesg -T | tail -30`.
3. **Narrow down to the service.** Collect:
- logs from `docker compose logs --tail=200`
- container state from `docker inspect` (`ExitCode`, `OOMKilled`, `RestartCount`, health)
- listening ports from `ss -tlnp`
- whether the service is reachable from the server itself and from the client.
If the app has a file under `../add-service/references/apps/`, read it.
4. **Form a hypothesis.** It must explain *all* the evidence. Confirm it with a read-only check before changing anything.
5. **Fix the cause.** Make the smallest change that addresses the cause rather than the symptom, following the fixed rules. Restarting a crash-looping container is not a fix.
6. **Verify.** Rerun the step 1 reproduction; it must now pass.
**Done when:** the root cause is named along with the evidence that proves it, and the step 1 reproduction passes. If the cause stays unknown, the task is done when each ruled-out cause is listed with its evidence and the user has the next check to run.
## Common causes
Most homeserver incidents are one of these. Test each against the evidence.
- **Disk or inodes full:** databases corrupt and containers fail to start. Container logs (`/var/lib/docker/containers/*/*-json.log`), old images and the journal are the usual space hogs.
- **Out of memory:** `docker inspect` shows `OOMKilled: true`, or the kernel log has `oom` lines. Machine-learning containers and CPU transcodes are the usual culprits.
- **Permissions:** PUID/PGID or volume ownership no longer matches, often after data was moved or a new image changed the user it runs as.
- **Image update:** a `latest` tag pulled a breaking release. Compare the image's creation date with when the breakage started, then read the release notes.
- **Data mount missing:** the data disk didn't mount at boot, so the app writes to or reads an empty directory on the OS disk. Check `findmnt`, and see the mount guards in `../setup/references/storage.md`.
- **Port conflict:** two stacks, or a stack and a host service, are bound to the same port.
- **DNS:** a container can't resolve names, or the LAN's own DNS (AdGuard or Pi-hole) went down and took name resolution with it.
- **Dying drive:** I/O errors in `dmesg`, or SMART reallocated or pending sectors.
- **Address changed:** the server got a new DHCP lease, or its Tailscale node key expired.
- **Power or heat:** throttling, undervoltage on a Pi (`vcgencmd get_throttled`), or a laptop that went to sleep.
- **Expired TLS certificate** on a reverse proxy.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · alexzfe/hsops | 100 | MIT |
| markusleben/ha-nova | 100 | MIT |
| als-apg/osprey | 100 | BSD-3-Clause |
| DeevsDeevs/agent-system | 100 | Apache-2.0 |
| terrylica/cc-skills | 100 | MIT |
| afrotools/afrotools | 100 | Apache-2.0 |
| YuDefine/nuxt-supabase-starter | 100 | MIT |
| redpanda-data/ui-harness | 100 | MIT |
| Sellyew/Workbuddy-workflow | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki ff21d43afe5a (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100ssh-diagnose
Diagnozuje problemy zdalnego serwera, np. wolną pracę, awarię usługi, pełny dysk, obciążenie CPU, pamięci lub sieci. - 86monitoring
Dodaje lub rozbudowuje monitoring serwera domowego: kontrolę dostępności, alerty, pulpity, metryki i zdrowie dysków oraz kopii zapasowych. - 90server-health
Sprawdza kondycję zdalnego serwera administrowanego przez użytkownika: procesor, pamięć, dysk, obciążenie, usługi i aktualizacje bezpieczeństwa. - 100quick-review
Wykonuje szybki przegląd serwera domowego i sprawdza, czy coś teraz szwankuje, bez pełnego audytu. - 90workload
Wdraża, uruchamia, skaluje i diagnozuje kontenery, aplikacje, API, usługi i zadania na platformie Control Plane. - 96docker-diagnose
Diagnozuje kontenery Docker, które nie startują, kończą się błędem, brakuje im pamięci lub nie łączą się, i proponuje naprawy.