Load tests: what many people at once do to the server, which the browser tests cannot show #133
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
The browser tests drive one person in one browser against a handful of fixture feeds, one request at a time. Nothing measures what production sees: several people at once, a catalogue of 1,500 feeds, the player saving positions every few seconds while a scan writes, many listeners seeking through files, a flood of sign-in attempts. Slow queries, lock contention and a server that stops answering only show there. Wanted: k6 scripts run against a scratch daemon with a realistic catalogue, with latency and error thresholds that fail the run.
Done in
b43ba89: tests/load, four k6 scripts (browse, listening, media, signin) run by node tests/load/run.js against a scratch daemon with 1,500 generated feeds, with ipx status polled throughout; k6 comes from its apt repository via /src/install.sh. It found #135 (fixed, in 0.10.0), #137 and #138 (fixed), and #136 (left open). Running on Postgres is #139, left open. All four pass with budgets of about twice today's numbers.