Duckeighty doctor

duckeighty: doctor

Run the checks below in order. For each, run the exact command, judge the result, and note the single fix command if it fails. At the end, produce a compact table (check / status / fix) and point the user to the single next action.

This skill only diagnoses. Do not run setup-ca, setup-resolver, or up from here — direct the user to the duckeighty-init skill or the manual command.

When a fix command is ./duckeighty.sh …, the user needs to run it from their duckeighty checkout. Resolve the path in this order, no guessing:

1. $DUCKEIGHTY_CHECKOUT env var if set.

2. If the ingress is running, ask Docker directly:

   docker compose ls --format json \
     | python3 -c "import json,sys,os; p=next((x for x in json.load(sys.stdin) if x['Name']=='duckeighty'),None); print(os.path.dirname(os.path.dirname(p['ConfigFiles'])) if p else '')"
   

   This returns the checkout root (parent of infra/compose.yaml) when the duckeighty project is running.

3. Otherwise ask the user for the path.

If the user has no checkout at all, tell them to run the duckeighty-init skill first instead of printing a fix command.

1. Docker is usable

docker version --format '{{.Server.Version}}'

Fails → "Start Docker Desktop (or your Docker daemon) and re-run."

2. shared_ingress network exists

docker network inspect shared_ingress >/dev/null 2>&1 && echo ok || echo missing

missing → "Run ./duckeighty.sh up from the duckeighty checkout. It creates the network before compose up."

3. duckeighty compose project is up

docker compose ls --format json | python3 -c "import json,sys; print(next((p['Status'] for p in json.load(sys.stdin) if p['Name']=='duckeighty'),'missing'))"

Expect something like running(2). missing → ./duckeighty.sh up.

4. Both ingress containers exist and are on shared_ingress

docker ps --filter 'label=com.docker.compose.project=duckeighty' --format '{{.Names}}\t{{.Networks}}\t{{.Status}}'

Expect two rows: duckeighty-traefik-1 and duckeighty-dnsmasq-1, both Up. The traefik row's Networks column must contain shared_ingress; if not, the ingress compose diverged from upstream.

5. macOS resolver entry

test -f /etc/resolver/duckeighty.test && cat /etc/resolver/duckeighty.test

Expect:

nameserver 127.0.0.1
port 5354

Missing or wrong → "Run ./duckeighty.sh setup-resolver (sudo required, one-time)."

6. DNS actually answers

dig +short @127.0.0.1 -p 5354 probe.duckeighty.test

Expect 127.0.0.1. Empty or timeout → dnsmasq is not listening on 5354/udp on 127.0.0.1. Check docker logs <dnsmasq container>; confirm the published ports in infra/compose.yaml are intact.

7. Local CA is trusted

mkcert -CAROOT
ls "$(mkcert -CAROOT)/rootCA.pem" 2>/dev/null

Both must succeed. Missing → ./duckeighty.sh setup-ca. If mkcert itself is missing, ask the user to install it: brew install mkcert nss.

8. HTTPS path reaches Traefik

curl -sI https://no-such-app.duckeighty.test/ | head -1

A HTTP/2 404 response from Traefik is the happy path: DNS + TLS both work, and the hostname just has no matching router yet. Other failures:

• curl: (6) Could not resolve host → step 5 or 6 failed
• TLS / certificate errors → step 7 failed, or the user did not restart the browser / shell after installing the CA
• curl: (7) Failed to connect → step 3 failed, or ports 80/443 are occupied by another process

Report format

check                          status    fix
1. docker daemon               OK        —
2. shared_ingress network      OK        —
3. duckeighty compose project  OK        —
4. traefik + dnsmasq on net    OK        —
5. /etc/resolver/…             MISSING   ./duckeighty.sh setup-resolver
6. DNS resolution              FAIL      (after step 5)
7. mkcert local CA             OK        —
8. HTTPS reaches traefik       FAIL      (after step 5)

End with the single first command the user should run. Do not dump all per-step fixes when one upstream fix unblocks the rest.