Jaeger (Distributed Tracing System)

Overview

jaeger란? Uber Technologies에서 만든 분산형 추적시스템 오픈소스로서 마이크로 서비스 기반 분산 시스템 모니터링을 할수가 있다.

현재는 CNCF(Cloud Native Computing Foundation)의 프로젝트이다.

- 분산형 추적시스템에 대표적인 오픈소스로는 zipkin, jagger, appdash 가 대표적이다.

- 분산 컨텍스트 전파

- 분산 트랜잭션 모니터링

- 근본 원인 분석

- 서비스 의존성 분석

- 성능 / 대기 시간 최적화

Installation

client

application.yml

compile group: 'io.opentracing.contrib', name:'opentracing-spring-web-autoconfigure', version: '0.0.4'
compile group: 'com.uber.jaeger', name: 'jaeger-core', version: '0.27.0

---

jaegerConfig.java
 

import com.uber.jaeger.Configuration;

import com.uber.jaeger.samplers.ProbabilisticSampler;

import org.springframework.beans.factory.annotation.Value;

import org.springframework.boot.web.client.RestTemplateBuilder;

import org.springframework.context.annotation.Bean;

import org.springframework.web.client.RestTemplate;

@org.springframework.context.annotation.Configuration

public class JaegerConfig {

	@Value("trace.url")

	String JaggerUrl;

	@Bean

	public io.opentracing.Tracer jaggerTracer() {

		return new Configuration("ios", new
		Configuration.SamplerConfiguration(ProbabilisticSampler.TYPE, 1,
		JaggerUrl),
		new Configuration.ReporterConfiguration()).getTracer();
	}
}

--

Server

Helm chart

value.yaml
 

# Default values for jaeger.
# Default values for jaeger.
# This is a YAML-formatted file.
# Jaeger values are grouped by component. Cassandra values override subchart values
 
provisionDataStore:
  cassandra: false
  elasticsearch: true
 
tag: 1.9.0
 
storage:
  # allowed values (cassandra, elasticsearch)
  type: elasticsearch
  cassandra:
    host: cassandra
    port: 9042
    user: user
    password: password
  elasticsearch:
    scheme: http
    host: elasticsearch
    port: 9200
    user: elastic
    password: changeme
    nodesWanOnly: false
 
# Begin: Override values on the Cassandra subchart to customize for Jaeger
cassandra:
  image:
    tag: 3.11.3
  persistence:
    # To enable persistence, please see the documentation for the Cassandra chart
    enabled: false
  config:
    cluster_name: jaeger
    seed_size: 1
    dc_name: dc1
    rack_name: rack1
    endpoint_snitch: GossipingPropertyFileSnitch
# End: Override values on the Cassandra subchart to customize for Jaeger
 
# Begin: Default values for the various components of Jaeger
# This chart has been based on the Kubernetes integration found in the following repo:
# https://github.com/jaegertracing/jaeger-kubernetes/blob/master/production/jaeger-production-template.yml
#
# This is the jaeger-cassandra-schema Job which sets up the Cassandra schema for
# use by Jaeger
schema:
  annotations: {}
  image: jaegertracing/jaeger-cassandra-schema
  pullPolicy: IfNotPresent
  # Acceptable values are test and prod. Default is for production use.
  mode: prod
  resources: {}
    # limits:
    #   cpu: 500m
    #   memory: 512Mi
    # requests:
    #   cpu: 256m
    #   memory: 128Mi
  ## Additional pod labels
  ## ref: https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
  podLabels: {}
 
# Begin: Override values on the Elasticsearch subchart to customize for Jaeger
elasticsearch:
  image:
    tag: "5.4"
  cluster:
    name: "tracing"
  data:
    persistence:
      enabled: false
  rbac:
    create: false
 
agent:
  enabled: true
  annotations: {}
  image: jaegertracing/jaeger-agent
  pullPolicy: IfNotPresent
  collector:
    host: null
    port: null
  cmdlineParams: {}
  daemonset:
    useHostPort: false
  service:
    annotations: {}
    # List of IP ranges that are allowed to access the load balancer (if supported)
    loadBalancerSourceRanges: []
    type: LoadBalancer
    # zipkinThriftPort :accept zipkin.thrift over compact thrift protocol
    zipkinThriftPort: 5775
    # compactPort: accept jaeger.thrift over compact thrift protocol
    compactPort: 6831
    # binaryPort: accept jaeger.thrift over binary thrift protocol
    binaryPort: 6832
    # samplingPort: (HTTP) serve configs, sampling strategies
    samplingPort: 5778
  resources: {}
    # limits:
    #   cpu: 500m
    #   memory: 512Mi
    # requests:
    #   cpu: 256m
    #   memory: 128Mi
  nodeSelector: {}
  podAnnotations: {}
  ## Additional pod labels
  ## ref: https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
  podLabels: {}
  ## Allow the scheduling on tainted nodes (requires Kubernetes >= 1.6)
  ## Ref: https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
  tolerations: []
  useHostNetwork: false
  dnsPolicy: ClusterFirst
 
collector:
  enabled: true
  annotations: {}
  image: jaegertracing/jaeger-collector
  pullPolicy: IfNotPresent
  dnsPolicy: ClusterFirst
  cmdlineParams: {}
  replicaCount: 1
  service:
    annotations: {}
    # List of IP ranges that are allowed to access the load balancer (if supported)
    loadBalancerSourceRanges: []
    type: ClusterIP
    # tchannelPort: used by jaeger-agent to send spans in jaeger.thrift format
    tchannelPort: 14267
    # httpPort: can accept spans directly from clients in jaeger.thrift format
    httpPort: 14268
    # can accept Zipkin spans in JSON or Thrift
    zipkinPort: 9411
    healthCheckPort: 14269
  resources: {}
    # limits:
    #   cpu: 1
    #   memory: 1Gi
    # requests:
    #   cpu: 500m
    #   memory: 512Mi
  nodeSelector: {}
  podAnnotations: {}
  ## Additional pod labels
  ## ref: https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
  podLabels: {}
  ## Allow the scheduling on tainted nodes (requires Kubernetes >= 1.6)
  ## Ref: https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
  tolerations: []
 
query:
  enabled: true
  annotations: {}
  image: jaegertracing/jaeger-query
  pullPolicy: IfNotPresent
  dnsPolicy: ClusterFirst
  cmdlineParams: {}
  healthCheckPort: 16687
  basePath: /
  replicaCount: 1
  service:
    annotations: {}
    type: ClusterIP
    # List of IP ranges that are allowed to access the load balancer (if supported)
    loadBalancerSourceRanges: []
    # queryPort: externally accessible port for UI and API
    queryPort: 80
    # targetPort: the internal port the UI and API are exposed on
    targetPort: 16686
  ingress:
    enabled: false
    annotations: {}
    # Used to create an Ingress record.
    # hosts:
    #   - chart-example.local
    # annotations:
      # kubernetes.io/ingress.class: nginx
      # kubernetes.io/tls-acme: "true"
    # tls:
      # Secrets must be manually created in the namespace.
      # - secretName: chart-example-tls
      #   hosts:
      #     - chart-example.local
  resources: {}
    # limits:
    #   cpu: 500m
    #   memory: 512Mi
    # requests:
    #    cpu: 256m
    #    memory: 128Mi
  nodeSelector: {}
  podAnnotations: {}
  ## Additional pod labels
  ## ref: https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
  podLabels: {}
  ## Allow the scheduling on tainted nodes (requires Kubernetes >= 1.6)
  ## Ref: https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
  tolerations: []
 
spark:
  enabled: false
  annotations: {}
  image: jaegertracing/spark-dependencies
  tag: latest
  pullPolicy: Always
  schedule: "49 23 * * *"
  successfulJobsHistoryLimit: 5
  failedJobsHistoryLimit: 5
  resources: {}
    # limits:
    #   cpu: 500m
    #   memory: 512Mi
    # requests:
    #   cpu: 256m
    #   memory: 128Mi
  nodeSelector: {}
  ## Additional pod labels
  ## ref: https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
  podLabels: {}
  ## Allow the scheduling on tainted nodes (requires Kubernetes >= 1.6)
  ## Ref: https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
  tolerations: []
# End: Default values for the various components of Jaeger
 
hotrod:
  enabled: false
  replicaCount: 1
  image:
    repository: jaegertracing/example-hotrod
    tag: latest
    pullPolicy: Always
  service:
    annotations: {}
    name: hotrod
    type: ClusterIP
    # List of IP ranges that are allowed to access the load balancer (if supported)
    loadBalancerSourceRanges: []
    externalPort: 80
    internalPort: 8080
  ingress:
    enabled: false
    # Used to create Ingress record (should be used with service.type: ClusterIP).
    hosts:
      - chart-example.local
    annotations: {}
      # kubernetes.io/ingress.class: nginx
      # kubernetes.io/tls-acme: "true"
    tls:
      # Secrets must be manually created in the namespace.
      # - secretName: chart-example-tls
      #   hosts:
      #     - chart-example.local
  resources: {}
    # We usually recommend not to specify default resources and to leave this as a conscious
    # choice for the user. This also increases chances charts run on environments with little
    # resources, such as Minikube. If you do want to specify resources, uncomment the following
    # lines, adjust them as necessary, and remove the curly braces after 'resources:'.
    # limits:
    #   cpu: 100m
    #   memory: 128Mi
    # requests:
    #   cpu: 100m
    #   memory: 128Mi
  tracing:
    host: null
    port: 6831

local docker

docker run --rm -it --network=host jaegertracing/all-in-one

Zipikin vs Jaeger

Launguage

	Zipkin	Jaeger
C++	Unofficial	Official
C#	Official	Official
Go	Official	Official
Java	Official	Official
Python	Unofficial	Official
Ruby	Official	Unofficial
Scala	Official	Unofficial(or use the Java library)
Node.js	Official	Official
PHP	Unofficial	Unofficial

Framework & Library

zipkin의 경우 인기있는 프레임워크를 지원하고 데이터베이스 드라이버와 같은 라이브러리를 지원한다.

jagger의 경우 Open Tracinginstrumentation libraries 를 활용한 다양한 라이브러리로 구현이 가능하다

jager는 또 gRPC, Thrift 및 AWS SDK 도구가 포함되어있다.

둘다 백엔드는 카산드라와 엘라스틱서치를 지원한다.

Jaeger는 CNCF의 일부로 당연히 kubernetes환경에 적합하게 설계되어있다.

helm 차트 템플릿 과 Envoy 또는 Isito 와 같은 서비스 프록시를 Jaeger Suppoter와 함께 사용하면 컨테이너에서 호출을 추적하는 것이

용이하다.

즉 k8s 통합환경에서 분산 추적에 맞게 설계되어있다.

Zipkin은 Docker 이미지와 Java 라이브러리를 제공Jaeger와 달리 Zipkin은 수집기, 저장소, API 및 UI가 포함 된 단일 프레임워크이다.

Conclusion

production환경이 kubernetes면 무조건 jaeger를 단일 환경 및 모놀리틱 시스템에는 zipkin을 고려해볼만 하다

즉 컨테이너 환경에는 jaeger로 가야되는게 맞는거 같다.

또한 jaeger안에는 expoter로 zipkin도 있기때문에 zipkin도 통합할수 있다.

즉 jaeger를 써보고 맞지않는다면 zipkin을 고려해보면 좋겠다.