Skip to content

Revert the watchCycle delay to Milliseconds

Craig Miskell requested to merge fix-watch-cycle-units into master

https://gitlab.com/gitlab-org/charts/components/gitlab-logger/-/merge_requests/2 refactored some code and in the process changed the loop-cycle interval from 100 Milliseconds to 100 Microseconds, possibly inadvertently (there's no strong indication in the commits/comments that it was meant).

In practice this means a 1000-fold increase in the rate of looking for new files on disk, and empirically on production this is CPU-intensive and expensive, and may have contributed to a production incident. Given the usual lag of logging systems is measured in units of seconds, checking every 0.1s is quite reasonable, and checking every 0.0001s is entirely unnecessary.

Merge request reports

Loading