All Projects → 1pkg → Gohalt

1pkg / Gohalt

Licence: mit
Gohalt 👮‍♀🛑: Fast; Simple; Powerful; Go Throttler library

Programming Languages

go
31211 projects - #10 most used programming language
golang
3204 projects

Projects that are alternatives of or similar to Gohalt

Kenshin
Kenshin: A time-series database alternative to Graphite Whisper with 40x improvement in IOPS
Stars: ✭ 203 (-19.76%)
Mutual labels:  monitoring, metrics
Snmpcollector
A full featured Generic SNMP data collector with Web Administration Interface for InfluxDB
Stars: ✭ 216 (-14.62%)
Mutual labels:  monitoring, metrics
Javamelody
JavaMelody : monitoring of JavaEE applications
Stars: ✭ 2,486 (+882.61%)
Mutual labels:  monitoring, metrics
Icingaweb2 Module Grafana
Grafana module for Icinga Web 2 (supports InfluxDB & Graphite)
Stars: ✭ 190 (-24.9%)
Mutual labels:  monitoring, metrics
Prometheus rabbitmq exporter
Prometheus.io exporter as a RabbitMQ Managment Plugin plugin
Stars: ✭ 248 (-1.98%)
Mutual labels:  monitoring, metrics
Sdk
Library for using Grafana' structures in Go programs and client for Grafana REST API.
Stars: ✭ 193 (-23.72%)
Mutual labels:  monitoring, metrics
Docker Traefik Prometheus
A Docker Swarm Stack for monitoring Traefik with Promethues and Grafana
Stars: ✭ 215 (-15.02%)
Mutual labels:  monitoring, metrics
Collector
pganalyze statistics collector for gathering PostgreSQL metrics and log data
Stars: ✭ 181 (-28.46%)
Mutual labels:  monitoring, metrics
Github Exporter
Prometheus exporter for github metrics
Stars: ✭ 231 (-8.7%)
Mutual labels:  monitoring, metrics
Hawkular Metrics
Time Series Metrics Engine based on Cassandra
Stars: ✭ 225 (-11.07%)
Mutual labels:  monitoring, metrics
Bigquery Grafana
Google BigQuery Datasource Plugin for Grafana.
Stars: ✭ 188 (-25.69%)
Mutual labels:  monitoring, metrics
Mtail
extract internal monitoring data from application logs for collection in a timeseries database
Stars: ✭ 3,028 (+1096.84%)
Mutual labels:  monitoring, metrics
Tobs
tobs - The Observability Stack for Kubernetes. Easy install of a full observability stack into a k8s cluster with a CLI tool or Helm charts.
Stars: ✭ 186 (-26.48%)
Mutual labels:  monitoring, metrics
Sematext Agent Docker
Sematext Docker Agent - host + container metrics, logs & event collector
Stars: ✭ 194 (-23.32%)
Mutual labels:  monitoring, metrics
Myperf4j
High performance Java APM. Powered by ASM. Try it. Test it. If you feel its better, use it.
Stars: ✭ 2,281 (+801.58%)
Mutual labels:  monitoring, metrics
Prometheus Net.dotnetruntime
Exposes .NET core runtime metrics (GC, JIT, lock contention, thread pool) using the prometheus-net package
Stars: ✭ 214 (-15.42%)
Mutual labels:  monitoring, metrics
Opencensus Web
A stats collection and distributed tracing framework
Stars: ✭ 168 (-33.6%)
Mutual labels:  monitoring, metrics
Ostent
Ostent is a server tool to collect, display and report system metrics.
Stars: ✭ 171 (-32.41%)
Mutual labels:  monitoring, metrics
Graphite exporter
Server that accepts metrics via the Graphite protocol and exports them as Prometheus metrics
Stars: ✭ 217 (-14.23%)
Mutual labels:  monitoring, metrics
Opencensus Node
A stats collection and distributed tracing framework
Stars: ✭ 249 (-1.58%)
Mutual labels:  monitoring, metrics

gohalt

Gohalt 👮‍♀🛑: Fast; Simple; Powerful; Go Throttler library

lint test report version license godoc

go get -u github.com/1pkg/gohalt

Introduction

Gohalt is simple and convenient yet powerful and efficient throttling go library. Gohalt provides various throttlers and surronding tools to build throttling pipelines and rate limiters of any complexity adjusted to your specific needs. Gohalt provides an easy way to integrate throttling and rate limiting with your infrastructure through built in middlewares.

Features

  • [x] Blastly fast and efficient, Gohalt has minimal performance overhead, it was design with performance as the primary goal.

  • [x] Flexible and powerful, Gohalt supports numbers of different throttling strategies and conditions that could be easily combined and customized to match your needs link.

  • [x] Easy to integrate, Gohalt provides separate package with numbers of built in middlewares for simple (couple lines of code) integrations with stdlib and other libraries, among which are: io, rpc/grpc, http, sql, gin, etc.

  • [x] Metrics awareness, Gohalt could use Prometheus metrics as a conditions for throttling.

  • [x] Queueing and delayed processing, Gohalt supports throttling queueing which means you can easily save throttled query to rabbitmq/kafka stream to process it later.

  • [ ] Durable storage, Gohalt has embedded k/v storage to provide thtottling persistence and durability.

  • [ ] Meta awareness, Gohalt provides easy way to access inner throttlers state in form of meta that can be later exposed to logging, headers, etc.

Concepts

Gohalt uses Throttler as the core interface for all derived throttlers and surronding tools.

// Throttler defines core gohalt throttler abstraction and exposes pair of counterpart methods: `Acquire` and `Release`.
type Throttler interface {
	// Acquire takes a part of throttling quota or returns error if throttling quota is drained
	// it needs to be called right before shared resource acquire.
	Acquire(context.Context) error
	// Release puts a part of throttling quota back or returns error if this is not possible
	// it needs to be called just after shared resource release.
	Release(context.Context) error
}

Throttler interface exposes pair of counterpart methods: Acquire takes a part of throttling quota or returns error if throttling quota is drained and needs to be called right before shared resource acquire; Release puts a part of throttling quota back or returns error if this is not possible and needs to be called just after shared resource release; Note: all derived throttler implementations are thread safe, so they could be used concurrently without additional locking. Note: all acquired throttlers should be released exatly the same amount of times they have been acquired. Note: despite throttler Release method has the same signature as Acquire has, Release implementations should try to handle any internal error gracefully and return error back rarely, nevertheless all errors returned by Release should be handeled by client.

In Gohalt throtllers could be easily combined with each other to build complex pipelines. There are multiple composite throttlers (all, any, ring, pattern, not, generator, etc) as well as leaf throttlers (timed, latency, monitor, metric, percentile, etc) to work with in Gohalt. If you don't find in existing throttlers the one that fits your needs you can create custom throttler by implementing Throttler interface. Such custom throttler should work with existing Gohalt throttlers and tools out of box.

Gohalt includes multiple supporting surrounding tools to make throttling more sugary.

// Runnable defined by typical abstract async func signature.
// Runnable is used by `Runner` as a subject for execution.
type Runnable func(context.Context) error
// Runner defines abstraction to execute a set of `Runnable`
// and return possible execution error back.
// Runner is designed to simplify work with throttlers
// by managing `Acquire`/`Release` loop.
type Runner interface {
	// Run executes single prodived `Runnable` instance.
	Run(Runnable)
	// Result returns possible execution error back.
	Result() error
}

Runnable and Runner define slim abstraction for executable and executor in Gohalt. Runner insterface aims to provide similar interface as errgroup.Group does. So to run a single executable use Run to wait and get result use Result. There are two runners implementations in Gohalt:

  • sync func NewRunnerSync(ctx context.Context, thr Throttler) Runner
  • async func NewRunnerAsync(ctx context.Context, thr Throttler) Runner Both implementation accept throttler and context as input arguments and handle all throttling cycle internaly. This way client donesn't need to call neither Acquire nor Release manually, all this is done by the runner. This way the only thing that needs to be done to add throttling to existing code wrap existing executable by Runnable. The only difference between sync and async runner is that the async runner starts each new Runnable inside new goroutine and uses locks for its imternal state. Note: You can't use sync runner in async fashion with go syncr.Run(func(context.Context) error{}) this will cause data race, use async runner instead async.Run(func(context.Context) error{}).

Last but not least Gohalt uses context heavily inside and there are multiple helpers to provide data via context for throttles, see throttles list to know when to use them.

// WithTimestamp adds the provided timestamp to the provided context
// to determine latency between `Acquire` and `Release`.
// Resulted context is used by: `latency` and `percentile` throtttlers.
func WithTimestamp(ctx context.Context, ts time.Time) context.Context
// WithPriority adds the provided priority to the provided context
// to differ `Acquire` priority levels.
// Resulted context is used by: `priority` throtttler.
func WithPriority(ctx context.Context, priority uint8) context.Context
// WithKey adds the provided key to the provided context
// to add additional call identifier to context.
// Resulted context is used by: `pattern` and `generator` throtttlers.
func WithKey(ctx context.Context, key string) context.Context
// WithMessage adds the provided message to the provided context
// to add additional message that need to be used to context.
// Resulted context is used by: `enqueue` throtttler.
func WithMessage(ctx context.Context, message interface{}) context.Context
// WithMarshaler adds the provided marshaler to the provided context
// to add additional message marshaler that need to be used to context.
// Resulted context is used by: `enqueue` throtttler.
// Used in pair with `WithMessage`.
func WithMarshaler(ctx context.Context, mrsh Marshaler) context.Context
// WithParams facade call that respectively calls:
// - `WithTimestamp`
// - `WithPriority`
// - `WithKey`
// - `WithMessage`
// - `WithMarshaler`
func WithParams(ctx context.Context, ts time.Time, priority uint8, key string, message interface{}, marshaler Marshaler) context.Context

Also there is yet another throttling sugar func WithThrottler(ctx context.Context, thr Throttler, freq time.Duration) context.Context related to context. Which defines context implementation that uses parrent context plus throttler internally. Using it you can keep typical context patterns for cancelation handling and apply and combine it with throttling.

select {
	case <-ctx.Done():
		return ctx.Error()
	default:
}

If internal context throttler is throttling context done chanel will be closed respectively. Note such behavior is implemented by throttler long pooling with the specified frequency, so efficiently there will be additional throttling user in form of long pooling goroutine.

// complex throttler example
thr := NewThrottlerAll( // throttles only if all children throttle
	NewThrottlerPattern(
		Pattern{ // use throttler only if provided key matches `192.*.*.*` submask
			Pattern: regexp.MustCompile(`192\.[0-9]+\.[0-9]+\.[0-9]+`),
			Throttler: NewThrottlerAny( // throttles if any children throttles
				// throttles only if latency is above 50 millisecond
				NewThrottlerLatency(50*time.Millisecond, 5*time.Second),
				// throttles only if cpu usage is above 70%
				NewThrottlerMonitor(NewMonitorSystem(time.Minute), Stats{CPUUsage: 0.7}),
			),
		},
	),
	// throttles each not 3rd call
	NewThrottlerNot(NewThrottlerEach(3)),
	// enqueues provided message to queue
	NewThrottlerEnqueue(NewEnqueuerRabbit("amqp://user:[email protected]:5672/vhost", "queue", time.Minute)),
)

In gohalt v0.4.0 breaking change is introduced to replace all untyped errors with two major error types:

  • ErrorThreshold which defines error type that occurs if throttler reaches specified threshold.
  • ErrorInternal which defines error type that occurs if throttler internal error happens. You can find list of returning error types for all existing throttlers in throttlers table bellow or in documentation.
    Note: not every gohalt throttler must return error; some throttlers might cause different side effects like logging or call to time.Sleep instead.

Throttlers

Throttler Definition Description
echo func NewThrottlerEcho(err error) Throttler Always throttles with the specified error back.
- could return any specified error;
wait func NewThrottlerWait(duration time.Duration) Throttler Always waits for the specified duration.
square func NewThrottlerSquare(duration time.Duration, limit time.Duration, reset bool) Throttler Always waits for square growing [1, 4, 9, 16, ...] multiplier on the specified initial duration, up until the specified duration limit is reached.
If reset is set then after throttler riches the specified duration limit next multiplier value will be reseted.
jitter func NewThrottlerJitter(initial time.Duration, limit time.Duration, reset bool, jitter float64) Throttler Waits accordingly to undelying square throttler but also adds the provided jitter delta distribution on top.
Jitter value is normalized to [0.0, 1.0] range and defines which part of square delay could be randomized in percents.
Implementation uses secure crypto/rand as PRNG function.
context func NewThrottlerContext() Throttler Always throttless on done context.
- could return ErrorInternal;
panic func NewThrottlerPanic() Throttler Always panics with ErrorInternal.
each func NewThrottlerEach(threshold uint64) Throttler Throttles each periodic i-th call defined by the specified threshold.
- could return ErrorThreshold;
before func NewThrottlerBefore(threshold uint64) Throttler Throttles each call below the i-th call defined by the specified threshold.
- could return ErrorThreshold;
after func NewThrottlerAfter(threshold uint64) Throttler Throttles each call after the i-th call defined by the specified threshold.
- could return ErrorThreshold;
past func NewThrottlerPast(threshold time.Time) Throttler Throttles each call befor timestamp defined by the specified UTC time threshold.
- could return ErrorThreshold;
future func NewThrottlerFuture(threshold time.Time) Throttler Throttles each call after timestamp defined by the specified UTC time threshold.
- could return ErrorThreshold;
chance func NewThrottlerChance(threshold float64) Throttler Throttles each call with the chance p defined by the specified threshold.
Chance value is normalized to [0.0, 1.0] range.
Implementation uses secure crypto/rand as PRNG function.
- could return ErrorThreshold;
running func NewThrottlerRunning(threshold uint64) Throttler Throttles each call which exeeds the running quota acquired - release q defined by the specified threshold.
- could return ErrorThreshold;
buffered func NewThrottlerBuffered(threshold uint64) Throttler Waits on call which exeeds the running quota acquired - release q defined by the specified threshold until the running quota is available again.
priority func NewThrottlerPriority(threshold uint64, levels uint8) Throttler Waits on call which exeeds the running quota acquired - release q defined by the specified threshold until the running quota is available again.
Running quota is not equally distributed between n levels of priority defined by the specified levels.
Use func WithPriority(ctx context.Context, priority uint8) context.Context to override context call priority, 1 by default.
timed func NewThrottlerTimed(threshold uint64, interval time.Duration, quantum time.Duration) Throttler Throttles each call which exeeds the running quota acquired - release q defined by the specified threshold in the specified interval.
Periodically each specified interval the running quota number is reseted.
If quantum is set then quantum will be used instead of interval to provide the running quota delta updates.
- could return ErrorThreshold;
latency func NewThrottlerLatency(threshold time.Duration, retention time.Duration) Throttler Throttles each call after the call latency l defined by the specified threshold was exeeded once.
If retention is set then throttler state will be reseted after retention duration.
Use func WithTimestamp(ctx context.Context, ts time.Time) context.Context to specify running duration between throttler acquire and release.
- could return ErrorThreshold;
percentile func NewThrottlerPercentile(threshold time.Duration, capacity uint8, percentile float64, retention time.Duration) Throttler Throttles each call after the call latency l defined by the specified threshold was exeeded once considering the specified percentile.
Percentile values are kept in bounded buffer with capacity c defined by the specified capacity.
If retention is set then throttler state will be reseted after retention duration.
Use func WithTimestamp(ctx context.Context, ts time.Time) context.Context to specify running duration between throttler acquire and release.
- could return ErrorThreshold;
monitor func NewThrottlerMonitor(mnt Monitor, threshold Stats) Throttler Throttles call if any of the stats returned by provided monitor exceeds any of the stats defined by the specified threshold or if any internal error occurred.
Builtin Monitor implementations come with stats caching by default.
Use builtin NewMonitorSystem to create go system monitor instance.
- could return ErrorInternal;
- could return ErrorThreshold;
metric func NewThrottlerMetric(mtc Metric) Throttler Throttles call if boolean metric defined by the specified boolean metric is reached or if any internal error occurred.
Builtin Metric implementations come with boolean metric caching by default.
Use builtin NewMetricPrometheus to create Prometheus metric instance.
- could return ErrorInternal;
- could return ErrorThreshold;
enqueuer func NewThrottlerEnqueue(enq Enqueuer) Throttler Always enqueues message to the specified queue throttles only if any internal error occurred.
Use func WithMessage(ctx context.Context, message interface{}) context.Context to specify context message for enqueued message and func WithMarshaler(ctx context.Context, mrsh Marshaler) context.Context to specify context message marshaler.
Builtin Enqueuer implementations come with connection reuse and retries by default.
Use builtin func NewEnqueuerRabbit(url string, queue string, retries uint64) Enqueuer to create RabbitMQ enqueuer instance or func NewEnqueuerKafka(net string, url string, topic string, retries uint64) Enqueuer to create Kafka enqueuer instance.
- could return ErrorInternal;
adaptive func NewThrottlerAdaptive(threshold uint64, interval time.Duration, quantum time.Duration, step uint64, thr Throttler) Throttler Throttles each call which exeeds the running quota acquired - release q defined by the specified threshold in the specified interval.
Periodically each specified interval the running quota number is reseted.
If quantum is set then quantum will be used instead of interval to provide the running quota delta updates.
Provided adapted throttler adjusts the running quota of adapter throttler by changing the value by d defined by the specified step, it subtracts d^2 from the running quota if adapted throttler throttles or adds d to the running quota if it doesn't.
- could return ErrorThreshold;
pattern func NewThrottlerPattern(patterns ...Pattern) Throttler Throttles if matching throttler from provided patterns throttles.
Use func WithKey(ctx context.Context, key string) context.Context to specify key for regexp pattern throttler matching.
Pattern defines a pair of regexp and related throttler.
- could return ErrorInternal;
- could return any underlying throttler error;
ring func NewThrottlerRing(thrs ...Throttler) Throttler Throttles if the i-th call throttler from provided list throttle.
- could return ErrorInternal;
- could return any underlying throttler error;
all func NewThrottlerAll(thrs ...Throttler) Throttler Throttles call if all provided throttlers throttle.
- could return ErrorInternal;
any func NewThrottlerAny(thrs ...Throttler) Throttler Throttles call if any of provided throttlers throttle.
- could return ErrorInternal;
not func NewThrottlerNot(thr Throttler) Throttler Throttles call if provided throttler doesn't throttle.
- could return ErrorInternal;
suppress func NewThrottlerSuppress(thr Throttler) Throttler Suppresses provided throttler to never throttle.
retry func NewThrottlerRetry(thr Throttler, retries uint64) Throttler Retries provided throttler error up until the provided retries threshold.
If provided onthreshold flag is set even ErrorThreshold errors will be retried.
Internally retry uses square throttler with DefaultRetriedDuration initial duration.
- could return any underlying throttler error;
cache func NewThrottlerCache(thr Throttler, cache time.Duration) Throttler Caches provided throttler calls for the provided cache duration, throttler release resulting resets cache.
Only non throttling calls are cached for the provided cache duration.
- could return any underlying throttler error;
generator func NewThrottlerGenerator(gen Generator, capacity uint64, eviction float64) Throttler Creates new throttler instance that throttles if found key matching throttler throttles.
If no key matching throttler has been found generator used insted to provide new throttler that will be added to existing throttlers map.
Generated throttlers are kept in bounded map with capacity c defined by the specified capacity and eviction rate e defined by specified eviction value is normalized to [0.0, 1.0], where eviction rate affects number of throttlers that will be removed from the map after bounds overflow.
Use WithKey to specify key for throttler matching and generation.
- could return ErrorInternal;
- could return any underlying throttler error;

Integrations

Note: in gohalt v0.3.0 all integrations were moved to separate repository to make base gohalt repository dependencies footprint small.

go get -u github.com/1pkg/gohaltlib

Library Adapter
gin func NewMiddlewareGin(thr Throttler, with GinWith, on GinOn) gin.HandlerFunc
stdlib http handler func NewMiddlewareStd(h http.Handler, thr Throttler, with StdWith, on StdOn) http.Handler
echo func NewMiddlewareEcho(thr Throttler, with EchoWith, on EchoOn) echo.MiddlewareFunc
beego func NewMiddlewareBeego(thr Throttler, with BeegoWith, on BeegoOn) beego.FilterFunc
kit func NewMiddlewareKit(thr Throttler, with KitWith, on KitOn) endpoint.Middleware
mux func NewMiddlewareMux(h http.Handler, thr Throttler, with MuxWith, on MuxOn) http.Handler
httprouter func NewMiddlewareRouter(h http.Handler, thr Throttler, with RouterWith, on RouterOn) http.Handler
reveal func NewMiddlewareRevel(thr Throttler, with RevealWith, on RevealOn) revel.Filter
iris func NewMiddlewareIris(thr Throttler, with IrisWith, on IrisOn) iris.Handler
fasthttp func NewMiddlewareFast(h fasthttp.RequestHandler, thr Throttler, with FastWith, on FastOn) fasthttp.RequestHandler
stdlib rt func NewRoundTripperStd(rt http.RoundTripper, thr Throttler, with RoundTripperStdWith, on RoundTripperStdOn) http.RoundTripper
fasthttp rt func NewRoundTripperFast(rt RoundTripperFast, thr Throttler, with RoundTripperFastWith, on RoundTripperFastOn) RoundTripperFast
stdlib rpc client coded func NewRPCClientCodec(cc rpc.ClientCodec, thr Throttler, with RPCCodecWith, on RPCCodecOn) rpc.ClientCodec
stdlib rpc server coded func NewRPCServerCodec(sc rpc.ServerCodec, thr Throttler, with RPCCodecWith, on RPCCodecOn) rpc.ServerCodec
grpc client stream func NewGRPCClientStream(cs grpc.ClientStream, thr Throttler, with GRPCStreamWith, on GRPCStreamOn) grpc.ClientStream
grpc server stream func NewGrpServerStream(ss grpc.ServerStream, thr Throttler, with GRPCStreamWith, on GRPCStreamOn) grpc.ServerStream
go-micro client func NewMicroClient(thr Throttler, with MicroClientWith, on MicroOn) client.Wrapper
go-micro server func NewMicroHandler(thr Throttler, with MicroServerWith, on MicroOn) server.HandlerWrapper
stdlib net conn func NewNetConn(conn net.Conn, thr Throttler, with NetConnWith, on NetConnOn, mode NetConnMode) net.Conn
stdlib sql func NewSQLClient(cli SQLClient, thr Throttler, with SQLClientWith, on SQLClientOn) SQLClient
stdlib io reader func NewReader(r io.Reader, thr Throttler, with RWWith, on RWOn) io.Reader
stdlib io writer func NewWriter(w io.Writer, thr Throttler, with RWWith, on RWOn) io.Writer

Licence

Gohalt is licensed under the MIT License.
See LICENSE for the full license text.

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].