Add a new SLI for the internal API to the pages service that uses `gitlab_pages_domains_source_api_call_duration_bucket` for apdex and `gitlab_pages_domains_source_api_requests_total` for error rate.
From this comment:
Add a new SLI for the internal API to the pages service that uses gitlab_pages_domains_source_api_call_duration_bucket
for apdex and gitlab_pages_domains_source_api_requests_total
for error rate. Hopefully we can set the thresholds here tighter than the 1s
and 10s
we have on the Rails side already. Ideally they'd also be tighter than the thresholds set for 1. because this is part of that duration.
If we add this, it could also be interesting to add a internal_api_duration_s
to the structured logs, which would make debugging easier.
This information would then also be a proxy for cache-hits v. misses.