Sprite u-2t39cz3r stuck: wake gives 502, restart API returns 409

Questions / Help
1

Sprite: u-2t39cz3r (org cory-nezin, id sprite-61420f98-9bce-4252-b4ed-ded642aacfd6)
URL: https://u-2t39cz3r-b5wj7.sprites.app

Since about 2026-11-04 it no longer wakes. Status is “cold”.

  • HTTPS: 502 after ~38 s, or no response in 30 s (fly-request-id 01M6MDBVF9PAXS2NVQYA2Z0C7J-sjc)
  • sprite exec hangs; /checkpoints, /exec, /services time out (request_id 57819a0e-3f24-43b5-abec-75982272500c)
  • POST /v1/sprites/u-2t39cz3r/restart returns 409 “upstream restart conflict” every time (retryable: true), e.g. x-request-id GOT7H2QWRcYa8c4NEs2E
  • The dashboard Restart button gives the same error (23:30 UTC).
  • Another Sprite in the same org works normally.

Last change we made: around 21:01 UTC we stopped tailscaled and ran tailscaled --cleanup, after a few minutes of running it with TUN. It worked after that.

There are no backups we can reach and the disk holds important data, so please recover it without losing the filesystem. Could you recreate the underlying Machine, or tell us what is blocking restart?

2

Update: the Sprite came back by itself at 23:38 UTC, about 8 minutes after the last refused restart, and the data looks intact. No action is needed to recover it. Could you tell us what happened, why restart returned 409, and whether stopping tailscaled with kernel TUN and running tailscaled --cleanup could have caused it? We want to avoid a repeat.

3

Sorry for the disruption this caused, your Sprite’s machine is pinned to a specific host and that host briefly ran out of free CPU to accommodate the restart. A transient capacity issue on our side, reported generically as a 409 conflict. It resolved itself once CPU freed up and since this was only stop/start, your volume was never touched, so data is intact.

The tailscaled change ran inside your sprite’s guest sandbox, it didn’t cause this.