Idea #14645
Updated by Peter Amstutz over 5 years ago
Record Report client latency for each FUSE operation in arv-mount. on keep requests. GET average latency (across all keepstores; parameterized by keepstore) (if possible, time to first byte) PUT average latency (across all keepstores; parameterized by keepstore) GET average transfer speed (across all keepstores; parameterized by keepstore) PUT average transfer speed (across all keepstores; parameterized by keepstore) Average over: last 30 seconds, last 5 minutes, since start. Use @prometheus_client@ Python library to aggregate counters/averages, read them back out and report them in current crunchstat format. Report operation time and operation count totals and interval Because the counters are in crunchstat. @arvados.keep.KeepClient@ this means @prometheus_client@ will become a dependency of @arvados-python-client@.