Evaluate endpoint `Projects::NotesController#index` low urgency, given high traffic share (8%)
Part of #2026 (closed)
The endpoint currently has an urgency set to low
(target duration 1s), and performs really well at that urgency, but it doesn't yet meet the medium
target (0.5s).
Count | Over current default target 5s | Over medium target 0.5s |
---|---|---|
38,000,858 | 2,555 | 342,161 |
0.9999327647 | 0.9909959665 |
(source: https://log.gprd.gitlab.net/goto/ec930220-9718-11ed-85ed-e7557b0a598c)
So 99.99% of requests are faster than 5s, and 99.09% of requests are faster than 0.5s. So we need to try and improve 0.5% of requests if we want to set the target to medium without negatively affecting apdex-score.
During this issue we've discovered that the endpoint is using etag-caching. However, the attributes urgency
and target_duration_s
aren't correctly included in the logs. This caused the reports not to be correct and we were judging requests differently for the error budget.
For error budgets, these requests would have a target duration of 5s when there was a cache miss, and a target of 1s when there was a miss. This means that there isn't likely to be a big change by correcting this. The cache-hits would easily be below 1s.
When we've corrected the logging information. We can easily see what the current performance of the endpoint is.