Powder doser camera #5

Activity

1. sgbaird commented on Jul 29, 2026

   

   sgbaird

   on Jul 29, 2026

   MemberAuthor

   More actions

   Thanks @seth-leavitt for setting up the software and @me-madsen for taking downstairs

   
2. sgbaird commented on Oct 23, 2026

   

   sgbaird

   on Oct 23, 2026

   MemberAuthor

   More actions

   @claude is this one still online?
3. sgbaird commented on Oct 23, 2026

   

   sgbaird

   on Oct 23, 2026

   MemberAuthor

   More actions

   Cc @swcharles @williamulbz
4. claude commented on Oct 23, 2026

   

   claudeBot

   on Oct 23, 2026 – with Claude

   Last edited by claude

   More actions

   Claude finished @sgbaird's task in 9m 59s —— View job

   ---

   No, the powder-doser camera is offline

   The powder-doser camera (picam-d1pr) has been offline since Wed 10-21 at 14:59 MDT, about 31 h ago (it's now 22:14 MDT on 10-22). Someone moved the camera by hand, and the Pi went offline about a minute later. This doesn't look like the Wi-Fi fault from #9.

   • Read the repo context (pi/README.md, uptime/, workflow config)
   • Check YouTube for a live powder-doser stream
   • Check the Pi over Tailscale and on the campus Wi-Fi
   • Look at the last minutes of its final broadcast

   Evidence

   • Tailscale: rpi-zero2w-stream-cam-d1pr is offline. It was last seen 2026-10-21 at 14:59:33 MDT.
   • YouTube: the last powder-doser broadcast, Jox0gGueJWY, ran from 13:01 to 14:59:16 on 10-21. There has been no powder-doser broadcast since, so it missed the 21:00, 05:00, 13:00 and 21:00 restarts. The other two streams (atomizer / picam-ot2 and OT2 overhead / hqcam-ot2-lid) are live right now.
   • Campus Wi-Fi: I pinged from another lab Pi on the same subnet. The doser Pi's usual address answers neither ping nor ARP. The OT-2 Pi's address answers immediately, so devices on that network can normally reach each other.
   • The last two minutes of video (jump to 14:57:40):

   time (MDT, from the overlay)	what the camera shows
   until 14:57:39	the usual view of the doser on the balance
   14:57:40	a hand grabs the camera's pole
   14:57:51	the camera is lifted and swung away
   14:57:55–14:58:43	it points at a blank light-gray surface, with the stand in the frame
   14:58:44–14:58:51	someone reaches in next to the stand
   14:58:52	last frame

   After that, YouTube saw the feed stop at 14:59:02, the broadcast was ended at 14:59:16, and the Pi dropped off Tailscale at 14:59:33.

   What this means

   A powered Pi within range of byu-devices would have come back on its own. Tailscale would reconnect, and the scheduled reboots plus the #9 watchdog would restart the stream. It hasn't done either in 31 h, so it's most likely unplugged, or moved somewhere outside Wi-Fi coverage.

   To bring it back: plug the Pi Zero back in with the camera attached, somewhere it can see the doser. The stream starts on boot, and a new "powder doser stream picam-d1pr" broadcast should appear within about a minute. If it was taken down on purpose (for example, while the doser is being worked on), nothing else needs doing. Tag @claude once it's plugged in and I'll check that it's streaming.

   Notes

   • I can't tell what ended the broadcast at 14:59:16. It was either a Lambda end call or someone in YouTube Studio, and the AWS keys here can't read the Lambda's logs. The Pi keeps its journal across reboots, so it will show what happened once the Pi is back online.
   • All checks were read-only: Tailscale status, the YouTube API, and a ping/ARP check from the other lab Pi. YouTube blocks this runner, so I downloaded the last 2.5 min of the video (3 MB at 360p) through that Pi. I didn't change anything on either Pi or in the repo, so there's no branch or PR.