1 unstable release
0.1.0 | Jun 27, 2024 |
---|
#430 in Concurrency
35KB
553 lines
statman
statman: A global manager of atomic statistics
Stats Manager
Individual statistics are managed by a Stats
instance.
Most use-cases will utilize the global instance by calling Stats::global
,
however multiple Stats
managers may be instantiated using [Stats::new()
].
Individual stats are managed by name. Stats of the same name and type will update the same underyling value.
Stat impls
All stats are produced in an [Arc
] so they may be written to by multiple thread.
Counter
: A counter backed by anstd::sync::atomic::AtomicU64
GaugeF64
: A gauge backed by anatomic_float::AtomicF64
GaugeI64
: A gauge backed by anstd::sync::atomic::AtomicI64
Histogram
: Anhdrhistogram
encapsulated with aspin
lock.Set
: A set of strings backed by anscc::HashSet
Scheduler
Stats
is a manager of named stats and does not do any scheduling by default.
Users may elect to schedule their own calls to Stats::flush
or use the scheduler
module.
Resetting Stats
By default, stats do not reset between calls to flush.
This behavior can be overwritten for individual stats using set_reset_on_flush
.
It can be overwritten globally using Stats
set_default_reset_on_flush_for_*
functions.
Example
use std::{
sync::{
atomic::{AtomicBool, Ordering},
Arc,
},
thread,
time::{Duration, SystemTime},
};
use chrono::{DateTime, Utc};
use statman::{scheduler::FlushScheduler, snapshot::StatsSnapshot, Stats};
// signal handling for a graceful exit
let runflag = Arc::new(AtomicBool::new(true));
{
let runflag = Arc::clone(&runflag);
ctrlc::set_handler(move || {
runflag.store(false, Ordering::Relaxed);
})
.expect("Error setting Ctrl-C handler");
}
// create and spawn a scheduler that will print stats every 10 minutes
FlushScheduler::new(Stats::global())
.with_register_callback(|snapshot: &StatsSnapshot| {
println!(
"{} - {:?}",
DateTime::<Utc>::from(SystemTime::now()).format("%T"),
snapshot
)
})
.with_runflag(Arc::clone(&runflag))
.spawn(Duration::from_secs(10))
.unwrap();
// create a histogram stat that resets on every call to flush
let hist = Stats::global().histogram("myhist");
hist.set_reset_on_flush(true);
// record an ever increasing value to the histogram
let mut next_record = 0;
while runflag.load(Ordering::Relaxed) {
hist.record(next_record);
next_record += 1;
thread::sleep(Duration::from_millis(10));
}
Output:
22:11:00 - StatsSnapshot { counters: {}, gauges_f64: {}, gauges_i64: {}, histograms: {"myhist": HistogramSnapshot { count: 597, percentiles: [(50.0, 298), (90.0, 537), (99.9, 596), (99.99, 596), (100.0, 596)] }}, sets: {} }
22:11:10 - StatsSnapshot { counters: {}, gauges_f64: {}, gauges_i64: {}, histograms: {"myhist": HistogramSnapshot { count: 994, percentiles: [(50.0, 1093), (90.0, 1491), (99.9, 1590), (99.99, 1590), (100.0, 1590)] }}, sets: {} }
22:11:20 - StatsSnapshot { counters: {}, gauges_f64: {}, gauges_i64: {}, histograms: {"myhist": HistogramSnapshot { count: 994, percentiles: [(50.0, 2087), (90.0, 2485), (99.9, 2585), (99.99, 2585), (100.0, 2585)] }}, sets: {} }
License: MIT OR Apache-2.0
Dependencies
~3MB
~47K SLoC