| Ticket Hash: | ca410e6ac31233712213f85c63b8f0595a1e5aeb | ||
| Title: | Fossil server request children wedge in CLOSE_WAIT on aborted client connections -> site appears down for that client | ||
| Status: | Type: | ||
| Severity: | Priority: | ||
| Subsystem: | Resolution: | ||
| Last Modified: |
2026-09-20 14:43:52 51.7 days ago |
Created: |
2026-09-02 11:06:25 69.9 days ago |
| Version Found In: | |||
| Description: | ||||
| 2026-09-02 ~06:59 EDT: maintainer reported the server non-responsive. Five fossil server children (forked request handlers) sat ~5 min in state S with their client connections in CLOSE_WAIT and ~45KB stuck in Send-Q (client 76.139.x abandoned the requests; children never noticed). New connections were fast the whole time, so only that client saw an outage. Killing the five children cleared it instantly. Suspect trigger: heavy page renders (667-post Trunk Ops thread pages / ext console) slow enough that the browser times out mid-response. Needs: reproduce, then either bound heavy renders (thread pagination) or confirm fossil's client-abort handling / add a watchdog that reaps children stuck in CLOSE_WAIT >60s. Related: prompt-size ticket 9288f6cb (same oversized-thread root). | ||||