Egress completed automatically after 163s

Agents
,
1

Web Egress jobs against my project are ending with EGRESS_COMPLETE and error: "" at roughly 150–165 seconds, even when the captured page is healthy and RTMPS output to Mux is still active. Two reproducers:

  • Studio capture page: egress EG_XXLQNH7iAUHG, ended at ~163.4 s, no error. Source page reported alive at age 150 s, Twilio room in-progress with 2 connected participants and 1v/1a tracks.

  • Minimal isolation probe (single static HTML page with an animated canvas, no Twilio, no SDK, fresh Mux livestream created 2 s before egress start): egress EG_x5mzFxPzDADu, ended at ~155.5 s, no error. Page heartbeats arriving every 5 s up to age 150 s. Mux livestream was still active with reconnectWindow=600 after egress completed.

Both jobs use RTMPS to global-live.mux.com:443/app/``..., encoding preset H264_1080P_30, awaitStartSignal=false. No stopEgress, no updateStream was called from our side; LiveKit’s streamResults is empty in the terminal status. Please tell us why these egresses are ending COMPLETE.

2

Cc: @Debashis_Ray

In OSS egress, there’s exactly one normal-completion path for Web Egress aside from explicit StopEgress: a Chrome console value of END_RECORDING.
triggers a clean finalize >> EGRESS_COMPLETE with empty error. Exactly your symptom.

Two possibilities at the 150–165s mark:

(1) something on your page is logging that string, load the same page in local Chrome, watch DevTools console at the same elapsed time;
(2) Cloud has an undocumented duration cap on Web Egress, which the consistency across two unrelated pages strongly suggests.

Static probe reproduces it, so (2) is more likely. Worth pinging @CWilson with both EG_* IDs.