hundreds-father-404
02/24/2021, 9:18 PM--stats-collector-enable
--stats-aggregator-enable
--debug-stats-enable
enough-analyst-54434
02/24/2021, 9:32 PM-X
for historical context.hundreds-father-404
02/24/2021, 9:33 PMregister('-x', '--time', type=bool,
help='Output a timing report at the end of the run.')
happy-kitchen-89482
02/24/2021, 10:15 PMhundreds-father-404
02/24/2021, 10:17 PM--no-histograms
option (they're super chatty)happy-kitchen-89482
02/24/2021, 10:17 PMhappy-kitchen-89482
02/24/2021, 10:17 PMhundreds-father-404
02/24/2021, 10:25 PM--stats-show
or --exec-stats-show
? I get nervous claiming [stats]
hundreds-father-404
02/24/2021, 10:27 PM--exec-stats-log
?happy-kitchen-89482
02/24/2021, 10:27 PMhappy-kitchen-89482
02/24/2021, 10:27 PMhappy-kitchen-89482
02/24/2021, 10:28 PMhundreds-father-404
02/24/2021, 10:28 PM14:59:37.98 [INFO] Counters:
local_cache_read_errors: 0
local_cache_requests: 31
local_cache_requests_cached: 8
local_cache_requests_uncached: 23
local_cache_write_errors: 0
local_execution_requests: 23
remote_cache_read_errors: 0
remote_cache_requests: 0
remote_cache_requests_cached: 0
remote_cache_requests_uncached: 0
remote_cache_speculation_local_completed_first: 0
remote_cache_speculation_remote_completed_first: 0
remote_cache_write_errors: 0
remote_cache_write_finished: 0
remote_cache_write_started: 0
remote_execution_errors: 0
remote_execution_requests: 0
remote_execution_rpc_errors: 0
remote_execution_rpc_execute: 0
remote_execution_rpc_retries: 0
remote_execution_rpc_wait_execution: 0
remote_execution_success: 0
remote_execution_timeouts: 0
hundreds-father-404
02/24/2021, 10:28 PMI'd prefer we do a subsystem I think, in part so we can add things in the future like a --no-histograms option (they're super chatty)
hundreds-father-404
02/24/2021, 11:23 PMhappy-kitchen-89482
02/24/2021, 11:42 PMhappy-kitchen-89482
02/24/2021, 11:56 PMexec-stats
, I think just stats
is fine?hundreds-father-404
02/24/2021, 11:57 PM--stats
, which I'm fine with. But does that mess up something like [black].stats
? I don't recall the rules. (I'll test if you don't remember)hundreds-father-404
02/25/2021, 2:19 AM19:18:51.74 [INFO] Histogram for `local_store_read_blob_size`:
min: 10
max: 3069951
mean: 115912.684
std dev: 515510.627
p25: 79
p50: 91
p75: 166
p90: 1644
p95: 468479
p99: 3069951
Anything else pertinent like "num buckets"? cc @witty-crayon-22786fast-nail-55400
02/25/2021, 2:21 AMfast-nail-55400
02/25/2021, 2:21 AMfast-nail-55400
02/25/2021, 2:22 AMfast-nail-55400
02/25/2021, 2:22 AMfast-nail-55400
02/25/2021, 2:23 AMhundreds-father-404
02/25/2021, 2:24 AM19:24:27.52 [INFO] Summary of `local_store_read_blob_size` observation histogram:
min: 10
max: 3069951
mean: 115914.174
std dev: 515510.321
total observations: 190
p25: 79
p50: 89
p75: 166
p90: 1644
p95: 468479
p99: 3069951
witty-crayon-22786
02/25/2021, 2:37 AMhundreds-father-404
02/25/2021, 2:38 AMwitty-crayon-22786
02/25/2021, 2:38 AM19:24:27.52 [INFO] local_store_read_blob_size:p50: 89
witty-crayon-22786
02/25/2021, 2:39 AMhundreds-father-404
02/25/2021, 2:45 AMAnd would continue to vote for just inlining into the counter values with a compound key:I think that gets too noisy. We have 28 counters, each histogram has 11 values associated with it. So adding 1 histogram means now 39 values in a flat list, ~30% corresponding to only a single metric