metrics - karapace_http_requests_in_progress may not be decrementing in all cases #1107
Description
What happened?
I have spammed several endpoints with ab (Apache HTTP server benchmarking tool) to stimulate the service to produce variability in the metrics. It appears there are cases where the request should have long terminated or timed out, but the value of karapace_http_requests_in_progress did not go down to zero. Endpoints like /subject and /topics had persistent values long (days) after I called them.
What did you expect to happen?
The value of all metrics should go to zero except maybe the /metrics and /health endpoints because they are hit by kubernetes and prometheus on a schedule.
What else do we need to know?
See 28c37db#diff-c552e10e92a4e3b45de60263b769f4e67784b3587e5e60fd308a7835400c58c0R9