Hi, all.
I find the output of serverStatus
includes many latencies counter, which can be used by prometheus function to do different kinds of calculation such as getting the changing rate of the metrics by rate(counter[1m])
However many latencies counter aren’t paired with its counterpart, a counter metrics. For example, application threads page read from disk to cache time (usecs)
has been paired with application threads page read from disk to cache count
, so users can easily get how much time spent on application threads read from disk per reads per second by :
rate(mongodb_ss_wt_cache_application_threads_page_read_from_disk_to_cache_time_usecs{instance=~"$instance"}[30s]) / rate(mongodb_ss_wt_cache_application_threads_page_read_from_disk_to_cache_count{instance=~"$instance"}[30s])
That’s a quite useful derived metric when troubleshooting. However some latencies have no counter metrics such as
application thread time evicting (usecs)
get reference for page index and slot time sleeping
Is that necessary to add them ?