跳转至

Pyroscope

DataKit 从 Version-1.67.0 版本开始增加了 Pyroscope 采集器，支持接入 Grafana Pyroscope Agent 上报的数据，帮助用户定位应用程序中的 CPU、内存、IO 等的性能瓶颈。

采集器配置

进入 DataKit 安装目录下的 conf.d/pyroscope 目录，复制 pyroscope.conf.sample 并命名为 pyroscope.conf。配置文件说明如下：

[[inputs.pyroscope]]
  ## pyroscope Agent endpoints register by version respectively.
  ## Endpoints can be skipped listen by remove them from the list.
  ## Default value set as below. DO NOT MODIFY THESE ENDPOINTS if not necessary.
  endpoints = ["/ingest"]

  ## set true to enable election, pull mode only
  election = true

  ## the max allowed size of http request body (of MB), 32MB by default.
  body_size_limit_mb = 32 # MB

  ## set false to stop generating apm metrics from ddtrace output.
  generate_metrics = true

  ## io_config is used to control profiling uploading behavior.
  ## cache_path set the disk directory where temporarily cache profiling data.
  ## cache_capacity_mb specify the max storage space (in MiB) that profiling cache can use.
  ## clear_cache_on_start set whether we should clear all previous profiling cache on restarting Datakit.
  ## upload_workers set the count of profiling uploading workers.
  ## send_timeout specify the http timeout when uploading profiling data to dataway.
  ## send_retry_count set the max retry count when sending every profiling request.
  # [inputs.pyroscope.io_config]
  #   cache_path = "/usr/local/datakit/cache/pyroscope_inputs"  # C:\Program Files\datakit\cache\pyroscope_inputs by default on Windows
  #   cache_capacity_mb = 10240  # 10240MB
  #   clear_cache_on_start = false
  #   upload_workers = 8
  #   send_timeout = "75s"
  #   send_retry_count = 4

  ## set custom tags for profiling data
  # [inputs.pyroscope.tags]
  #   some_tag = "some_value"
  #   more_tag = "some_other_value"

配置好后，重启 DataKit ，开启 Pyroscope 采集器。

目前可以通过 ConfigMap 方式注入采集器配置来开启采集器。

设置采集器全局 Tag

可以在采集器配置中通过 [inputs.pyroscope.tags] 指定额外标签，该标签会统一应用到所有该采集器采到的数据：

 [inputs.pyroscope.tags]
  # some_tag = "some_value"
  # more_tag = "some_other_value"
  # ...

客户端 SDK 接入

Pyroscope 采集器目前支持 Java，Python， Go 和 Rust 等语言的 Pyroscope Agent 接入，其他语言正在持续支持中：

从 Github 下载最新的 pyroscope.jar 包，作为 Java Agent 启动你的应用：

PYROSCOPE_APPLICATION_NAME="java-pyro-demo" \
PYROSCOPE_LOG_LEVEL=debug \
PYROSCOPE_FORMAT="jfr" \
PYROSCOPE_PROFILER_EVENT="cpu" \
PYROSCOPE_LABELS="host=$(hostname),service=java-pyro-demo,version=1.2.3,env=dev,some_other_tag=other_value" \
PYROSCOPE_UPLOAD_INTERVAL="60s" \
PYROSCOPE_JAVA_STACK_DEPTH_MAX=512 \
PYROSCOPE_PROFILING_INTERVAL="10ms" \
PYROSCOPE_PROFILER_ALLOC=128k \
PYROSCOPE_PROFILER_LOCK=10ms \
PYROSCOPE_ALLOC_LIVE=false \
PYROSCOPE_GC_BEFORE_DUMP=true \
PYROSCOPE_SERVER_ADDRESS="http://127.0.0.1:9529" \
java -javaagent:pyroscope.jar -jar your-app.jar

更多细节请参考 Grafana 官方文档

  1. 安装 pyroscope-io 依赖包： 

    pip install pyroscope-io

  2. 代码引入 pyroscope-io 包： 

    import os
import pyroscope
import socket

pyroscope.configure(
    server_address="http://127.0.0.1:9529",
    detect_subprocesses=True,
    oncpu=True,
    enable_logging=True,
    report_pid=True,
    report_thread_id=True,
    report_thread_name=True,
    tags={
        "host": socket.gethostname(),
        "service": 'python-pyro-demo',
        "version": 'v1.2.3',
        "env": "testing",
        "process_id": os.getpid(),
    }
)

  3. 启动应用： 

    PYROSCOPE_APPLICATION_NAME="python-pyro-demo" python app.py

  1. 添加 pyroscope-go 模块： 

    go get github.com/grafana/pyroscope-go

  2. 引入模块并初启动： 

    import (
    "log"
    "os"
    "runtime"
    "strconv"
    "time"

    "github.com/grafana/pyroscope-go"
)

func Must[T any](t T, _ error) T {
    return t
}

runtime.SetMutexProfileFraction(5)
runtime.SetBlockProfileRate(5)

profiler, err := pyroscope.Start(pyroscope.Config{
    ApplicationName: "go-pyroscope-demo",

    // replace this with the address of pyroscope server
    ServerAddress: "http://127.0.0.1:9529",

    // you can disable logging by setting this to nil
    Logger: pyroscope.StandardLogger,

    // uploading interval period
    UploadRate: time.Minute,

    // you can provide static tags via a map:
    Tags: map[string]string{
        "service":    "go-pyroscope-demo",
        "env":        "demo",
        "version":    "1.2.3",
        "host":       Must(os.Hostname()),
        "process_id": strconv.Itoa(os.Getpid()),
        "runtime_id": UUID,
    },

    ProfileTypes: []pyroscope.ProfileType{
        // these profile types are enabled by default:
        pyroscope.ProfileCPU,
        pyroscope.ProfileAllocObjects,
        pyroscope.ProfileAllocSpace,
        pyroscope.ProfileInuseObjects,
        pyroscope.ProfileInuseSpace,

        // these profile types are optional:
        pyroscope.ProfileGoroutines,
        pyroscope.ProfileMutexCount,
        pyroscope.ProfileMutexDuration,
        pyroscope.ProfileBlockCount,
        pyroscope.ProfileBlockDuration,
    },
})
if err != nil {
    log.Fatal("unable to bootstrap pyroscope profiler: ", err)
}

defer profiler.Stop()

Note

Pyroscope Rust agent 目前只能正常工作在 Linux 平台上。

  1. 添加 pyroscope 和 pyroscope_pprofrs crates 到项目依赖中： 

    cargo add pyroscope
cargo add pyroscope_pprofrs

  2. 代码中初始化并启动 Pyroscope Rust profiling 任务： 

    use pyroscope::{PyroscopeAgent, Result};
use pyroscope_pprofrs::{pprof_backend, PprofConfig};

fn main() -> Result<()> {
    let pprof_config = PprofConfig::new().sample_rate(100).report_thread_id().report_thread_name(); // 采样率等基础配置
    let backend_impl = pprof_backend(pprof_config);

    // Pyroscope agent 配置
    let agent = PyroscopeAgent::builder("http://127.0.0.1:9529", "pyroscope-rust-app")
        .backend(backend_impl)
        .tags([("version", "1.23.4"), ("env", "demo"), ("host", "<your-hostname>")].to_vec())
        .build()?;

    // start the Pyroscope agent
    let agent_running = agent.start()?;

    // your application code...

    // gracefully shutdown the Pyroscope agent
    let agent_ready = agent_running.stop()?;
    agent_ready.shutdown();

    Ok(())
}

通过与链路数据之间的关联（Grafana 称之为 Span profiles），用户可以更容易的洞察到系统的性能瓶颈，Pyroscope 提供了相关的 OpenTelemetry 插件，可以让两者之间的数据关联起来，目前支持 Java，Python 和 Go 等语言，下面分别介绍。

Note

为了便于区分同一服务的不同实例，我们可以在进程启动时随机生成一个 UUID，然后在进程的整个生命周期内把该 UUID 作为 runtime_id 标签设置到所有的链路和 profiling 数据上，这样便能关联两者数据。 除了 runtime_id 标签，还建议所有应用添加 host（主机名），service（服务名），version（服务版本），env（部署环境）， process_id（服务启动进程号）等标签，方便关联各类采集到的指标和数据。

  1. 下载 OpenTelemetry 官方提供的 Java agent 包 opentelemetry-javaagent.jar。
  2. 下载 Pyroscope OpenTelemetry 插件 pyroscope-otel.jar。
  3. 进行相应设置并启动你的 Java 应用 
    uuid=$(uuidgen);
OTEL_SERVICE_NAME="java-pyro-demo" \
OTEL_RESOURCE_ATTRIBUTES="runtime_id=$uuid,host=$(hostname),service.name=java-pyro-demo,service.version=1.3.55,service.env=dev" \
OTEL_JAVAAGENT_EXTENSIONS=./pyroscope-otel.jar \
OTEL_PYROSCOPE_ADD_PROFILE_URL=false \
OTEL_PYROSCOPE_ADD_PROFILE_BASELINE_URL=false \
OTEL_PYROSCOPE_START_PROFILING=true \
OTEL_TRACES_EXPORTER=otlp \
OTEL_EXPORTER_OTLP_PROTOCOL="http/protobuf" \
OTEL_EXPORTER_OTLP_TRACES_ENDPOINT="http://127.0.0.1:9529/otel/v1/traces" \
OTEL_EXPORTER_OTLP_METRICS_ENDPOINT="http://127.0.0.1:9529/otel/v1/metrics" \
OTEL_EXPORTER_OTLP_LOGS_ENDPOINT="http://127.0.0.1:9529/otel/v1/logs" \
OTEL_EXPORTER_OTLP_COMPRESSION=gzip \
PYROSCOPE_APPLICATION_NAME="java-pyro-demo" \
PYROSCOPE_LOG_LEVEL=debug \
PYROSCOPE_FORMAT="jfr" \
PYROSCOPE_PROFILER_EVENT="cpu" \
PYROSCOPE_LABELS="runtime_id=$uuid,service=java-pyro-demo,version=1.2.3,env=dev,host=$(hostname),other_tag=other_value" \
PYROSCOPE_UPLOAD_INTERVAL="60s" \
PYROSCOPE_JAVA_STACK_DEPTH_MAX=512 \
PYROSCOPE_PROFILING_INTERVAL="10ms" \
PYROSCOPE_PROFILER_ALLOC=128k \
PYROSCOPE_PROFILER_LOCK=10ms \
PYROSCOPE_ALLOC_LIVE=false \
PYROSCOPE_GC_BEFORE_DUMP=true \
PYROSCOPE_SERVER_ADDRESS="http://127.0.0.1:9529" \
java -javaagent:opentelemetry-javaagent.jar -jar your-app.jar
Tips

上述使用 uuidgen 命令随机生成了一个 UUID，并通过环境变量 OTEL_RESOURCE_ATTRIBUTES 和 PYROSCOPE_LABELS 分别为链路和 profiling 设置 runtime_id tag，其它一些环境变量的设置仅供参考，请根据实际情况或参考官方文档酌情增删。

  1. 安装 pyroscope-otel 依赖库 

    pip install pyroscope-otel

  2. 引入 pyroscope-otel 和 opentelemetry 库并进行相应配置 

    import uuid
import socket
import os
import pyroscope

from opentelemetry import trace
from opentelemetry.sdk.resources import Resource
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from pyroscope.otel import PyroscopeSpanProcessor
from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter


UUID = uuid.uuid1() // 进程启动时随机生成一个 UUID
otelExporter = OTLPSpanExporter(endpoint='http://127.0.0.1:4317', insecure=True, timeout=30)

tracerProvider = TracerProvider(resource=Resource(attributes={
    "service.name": "python-pyro-demo",
    "service.version": "v3.5.7",
    "service.env": "dev",
    "host": socket.gethostname(),
    "process_id": os.getpid(),
    "runtime_id": str(UUID), // 为链路设置 runtime_id tag
}))

tracerProvider.add_span_processor(PyroscopeSpanProcessor())
tracerProvider.add_span_processor(BatchSpanProcessor(span_exporter=otelExporter, max_queue_size=100, max_export_batch_size=30))
trace.set_tracer_provider(tracerProvider)
tracer = trace.get_tracer("python-pyro-demo")

pyroscope.configure(
    server_address="http://127.0.0.1:9529",
    detect_subprocesses=True,
    oncpu=True,
    enable_logging=True,
    report_pid=True,
    report_thread_id=True,
    report_thread_name=True,
    tags={
        "runtime_id": str(UUID), // 为 profiling 设置 runtime_id tag
        "host": socket.gethostname(),
        "service": 'python-pyro-demo',
        "version": 'v0.2.3',
        "env": "testing",
        "process_id": os.getpid(),
    }
)


if __name__ == '__main__':
    // your app code

pyroscope.shutdown()

  1. 添加 pyroscope-go 库到项目依赖中 

    go get github.com/grafana/pyroscope-go

  2. 配置并启动 OpenTelemetry 和 Pyroscope 

    package main

import (
    "github.com/google/uuid"
    otelpyroscope "github.com/grafana/otel-profiling-go"
    "github.com/grafana/pyroscope-go"
    "go.opentelemetry.io/otel"
    "go.opentelemetry.io/otel/attribute"
    "go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp"
    "go.opentelemetry.io/otel/propagation"
    "go.opentelemetry.io/otel/sdk/resource"
    tracesdk "go.opentelemetry.io/otel/sdk/trace"
    "go.opentelemetry.io/otel/trace"
)

var (
    UUID       = uuid.NewString() // 生成一个全局的 UUID
    otelTracer trace.Tracer
)

func hostname() string {
    host, _ := os.Hostname()
    return host
}

func main() {
    otelExporter, err := otlptracehttp.New(context.Background(),
        otlptracehttp.WithEndpointURL("http://127.0.0.1:9529/otel/v1/traces"),
        otlptracehttp.WithInsecure(),
        otlptracehttp.WithTimeout(time.Second*15),
    )
    if err != nil {
        log.Fatal("unable to init otel tracing exporter: ", err)
    }
    tracerProvider := tracesdk.NewTracerProvider(tracesdk.WithBatcher(otelExporter,
        tracesdk.WithBatchTimeout(time.Second*3)),
        tracesdk.WithResource(resource.NewSchemaless(
            attribute.String("runtime_id", UUID), // 为链路设置 runtime_id tag
            attribute.String("service.name", "go-pyroscope-demo"),
            attribute.String("service.version", "v0.0.1"),
            attribute.String("service.env", "dev"),
            attribute.String("host", hostname()),
            attribute.String("process_id", strconv.Itoa(os.Getpid()))
        )),
    )
    defer tracerProvider.Shutdown(context.Background())

    otel.SetTracerProvider(otelpyroscope.NewTracerProvider(tracerProvider))
    otelTracer = otel.Tracer("go-pyroscope-demo")
    log.Printf("otel tracing started....\n")

    runtime.SetMutexProfileFraction(5)
    runtime.SetBlockProfileRate(5)

    profiler, err := pyroscope.Start(pyroscope.Config{
        ApplicationName: "go-pyroscope-demo",

        // replace this with the address of pyroscope server
        ServerAddress: "http://127.0.0.1:9529",

        // you can disable logging by setting this to nil
        Logger: pyroscope.StandardLogger,

        // uploading interval period
        UploadRate: time.Minute,

        // you can provide static tags via a map:
        Tags: map[string]string{
            "runtime_id": UUID, // 为 profiling 设置 runtime_id tag
            "env":        "demo",
            "version":    "0.0.1",
            "host":       hostname(),
            "process_id": strconv.Itoa(os.Getpid()),
        },

        ProfileTypes: []pyroscope.ProfileType{
            // these profile types are enabled by default:
            pyroscope.ProfileCPU,
            pyroscope.ProfileAllocObjects,
            pyroscope.ProfileAllocSpace,
            pyroscope.ProfileInuseObjects,
            pyroscope.ProfileInuseSpace,

            // these profile types are optional:
            pyroscope.ProfileGoroutines,
            pyroscope.ProfileMutexCount,
            pyroscope.ProfileMutexDuration,
            pyroscope.ProfileBlockCount,
            pyroscope.ProfileBlockDuration,
        },
    })
    if err != nil {
        log.Fatal("unable to bootstrap pyroscope profiler: ", err)
    }

    log.Printf("pyroscope profiler started....\n")
    defer profiler.Stop()

    // your app code...
}

文档评价

文档内容是否对您有帮助？ ×