- Dikirim
- 24 September 2026 pukul 03.21 UTC
- Penulis
- Kamo
- Commit
- 86cc3b5
Every jicofo logs in to prosody as the same full JID, focus@auth.meet.Meet/focus. The maxSurge rollout from 9da0998 runs two of them at once, and prosody answers a second login on one JID by kicking the first with a stream `conflict`, which Smack never reconnects from. On 2026-09-23 a `rollout restart` of prosody and jicofo together let the outgoing jicofo reconnect to the new prosody after the incoming one had logged in. It displaced the new pod and then exited, so there was no focus at all: jicofo knew no bridge, and every meeting failed to start. The tcpSocket probe on 8888 hid it. The REST server binds that port regardless of the XMPP session, so the pod sat Running 1/1 for the whole outage. The probes now require an ESTABLISHED socket to prosody:5222 (0x1466 in /proc/net/tcp). Checked against both states: exit 1 on the dead pod, exit 0 on the recovered one.