Troubleshooting

When you hit a problem, first check it against the symptoms in this chapter. If none match, contact us via "Getting help" at the end.

1. No video (Live keeps buffering)

Possible cause	How to verify	Fix
Wrong RTSP URL	Try the same URL in VLC	Change the URL and re-save; when ONVIF auto-discovery returns a wrong address, switch to manual RTSP
Wrong credentials	The web admin reports auth failure	Go to camera details and change the password; the ONVIF password ≠ the RTSP password
Low-latency channel blocked	DevTools → Network shows a connection error	Open UDP 23515 in the firewall; if you can't, it auto-switches to the stable channel (2–4s latency)
Camera uses a proprietary protocol (Tapo / Xiaomi)	The camera brand isn't in the Cameras list	Use a go2rtc bridge
The streaming gateway didn't start	`docker logs yunkan` for startup errors	Usually a port conflict; try restarting the container

2. Playback is black / the progress bar won't move

Recording isn't on: Camera details → Recording settings → enable
Disk full: df -h to check the partition holding data, then delete old recordings or shorten the retention days
Corrupted recording segment: find the matching data/recordings/<id>/<date>/<HH-MM-SS>.mp4 and try playing it in VLC; a corrupted segment loses at most 5 minutes
Wrong time: if the server and camera time zones are out of sync, the playback timeline is off by 8 hours; check with timedatectl

3. AI detection isn't working

Symptom	Cause	Fix
The events page stays empty	The global detection toggle is off	Settings → Detection → enable
One camera doesn't detect	That camera has a "detection failed" status	Camera details → "Clear detection error"
GPU at 0 usage	Wrong image variant (running the cpu image on a GPU box)	Pull the matching variant image and restart
Faces always "Stranger"	This person was never added / the threshold is too strict	Add the face / lower the face similarity threshold
Many false fall alerts	The descent-velocity threshold is too low	Raise the fall descent-velocity threshold a bit

4. The container keeps restarting

Use docker logs --tail 200 yunkan to see the last output. Common errors:

Missing config: in wizard mode but the health check fails → just finish the /setup wizard in the browser
Port in use: ss -tnlp | grep 23406, then stop the conflicting service or change the port
Hardware files not mounted: the bind-mount didn't include /etc/machine-id, so license validation crashes; add it per Installation
Unsupported architecture: SkyView publishes x86_64 images only and can't run on ARM machines (uname -m should output x86_64)

5. Uploading to 115 fails

Login expired (about 30 days): Settings → 115 → re-scan to log in
VIP expired: very large files need 115 VIP's instant-upload channel; non-VIP uses a rate-limited channel
Target folder deleted: recreate the folder in 115, then return to SkyView and re-select it
Unstable network: upload tasks have built-in retries; check docker logs yunkan for the specific error

6. Automation doesn't trigger

Quiet hours: Settings → Notifications → Quiet hours, check whether you're currently within that window
Cooldown: repeated events within 30 seconds are swallowed; lower the cooldown or use a different event type
Condition mismatch: the rule added "recognized person = Zhang San" but it was actually recognized as "Stranger", so the rule won't trigger; relax the condition
MQTT not connected: Settings → Integrations → MQTT to check the connection status, and verify the address / credentials

7. The wizard reappears after upgrading / data is gone

Jumping to /setup after an upgrade is almost always because the data directory wasn't mounted correctly — the container can't see the config file written last time, so it assumes a fresh install. Check:

bash

docker inspect yunkan | grep -A 2 Mounts
# You should see Source: /home/.../data, Destination: /app/data
# If the path changed / isn't mounted, docker rm the container and mount it correctly on the next docker run

Never go through the wizard again

Going through setup writes a new empty database; the original is still there but deactivated. Stop the container first, fix the mount, and restart — you'll go straight into normal mode.

8. How to read logs

bash

# Single-image deploy: all logs are mixed together
docker logs --tail 500 -f yunkan

# To see only detection-related:
docker logs yunkan 2>&1 | grep -i detection

# Multi-container deploy: each component is separate
docker compose logs -f api detection

Getting help

Support ticket: Customer portal → My Orders → Open a ticket, attach the output of docker logs --tail 200 yunkan, fastest follow-up
Email: support@yun-kan.com (reply within 24 hours)
WeChat group: after purchase, the "My Orders" page in the customer portal has a group QR code (paid users only)