Repository navigation
Listing or downloading the entries of a large round times out #640
Description
For large rounds, listing or downloading the entries fails: both GET /v1/admin/round/<id>/entries and /entries/download are killed by gunicorn's 30 s timeout.
Reproduce (montage-dev, 2026-10-02)
A round with 15,676 entries (Images from Wiki Loves Monuments 2026 in Russia): both requests ended after 30.0–30.5 s with SystemExit('1') from handle_abort in the exception log. A round with 427 entries downloads in well under a second.
Cause [concluded from the code]
get_round_entries() loads every RoundEntry, and to_export_dict() then lazily loads each entry's Entry and RoundSource: several queries per row. Same on master. Not memory-bound here, but time-bound.
Possible fix
One query that joins round_entries, entries and round_sources and streams the CSV.
Written with Claude (Claude Code) on behalf of @lgelauff.
Activity
mahmoud commentedon Oct 10, 2026
Fixed in #659 (merged as 8ac4cfd): the entries list and CSV download now run two queries per request instead of two per entry (perf/round-entries-export-640). File-list rounds still repeat their source params in every row (#639).