Closed jasonLaster closed 2 years ago
I think the tests are running correctly again. I'm not sure precisely what was broken but documenting what I saw and the resolution steps for posterity:
d livetest-gecko
initially failed because the browser wasn't found. It was looking in .replay-playwright-browsers
instead of .replay
. It should have been using .replay
based on my reading of the playwright installer code and the RECORD_REPLAY_DIRECTORY
wasn't set to override it to something else. My next suspicion was that we had an out of date dependency. Running npm ls --depth=0
in the backend
showed a lot of dependency errors so I ran npm ci
to wipe and reinstall the pinned versions. This resolved finding the browser directory.Going to look into adding more telemetry next so we have better insight into these failures.
Yeah a telemetry event {name: "live-test", success: true, testName: "airbnb" }
Opened #89 to improve error handling. We already have telemetry for runlive tests but we lacked the test info so I've added that in https://github.com/RecordReplay/backend/pull/3198.
I restarted runlive on mac and we're getting the example in honeycomb now but none of the tests are passing. Investigating ...
The playwright browser was out of date (by a couple months) so it was unable to authenticate. Merged the latest webreplay-release
into replay-playwright
, ran d deploy-playwright
and d release-playwright
on macstadium, killed all runlive processes and open replay browsers, and started them again via d livetest-gecko
and we're getting sample recordings again!
https://ui.honeycomb.io/replay/datasets/backend/result/er82JjoAcGy
The playwright test suite stopped running recently we should bring it back and ensure that we have telemetry wired up so we can be alerted in honeycomb if it stops running again
We have one telemetry event for when the live tests fail to create a recording
It'd be nice to also track attempts so that we can measure uptime and alert when the service is down and we stop trying