yidinghan / eak

elasticsearch apm-server kibana

EAK

• EAK
• Quick Run
• Remarks
• Resources
• Waterline
  • Example

elasticsearch apm-server kibana

Quick Run

wget https://raw.githubusercontent.com/yidinghan/eak/master/docker-compose.yml
docker-compose up -d

Go to localhost:5601, or what your machine IP

Wait for kibana to be available, then you are ready to try apm

Remarks

• Since v6.4, apm dashboard setup move to Kibana UI.

  

• with apm-agent-nodejs:v1.1.0

  • start method

  // change from
  require('elastic-apm-node/start');
  // to
  require('elastic-apm-node').start();

  • elastic-apm-node.js config change

  // change from
  appName: 'apm-reseach',
  // to
  serviceName: 'apm-reseach',

• with docker.elastic.co/kibana/kibana:6.2.2

  • New tab APM is ready to go

  

• with playdingnow/elastic-apm-server:1.5.0

  • drop Waterline from dashboard temporary

• with docker.elastic.co/kibana/kibana:6.1.1

  • A new tab named APM appears in the left sidebar

  

• with playdingnow/elastic-apm-server:1.4.3

  • The default apm-dashboards.json have been changed to customize data
  • One more visualization chart name is Waterline

  

• with playdingnow/elastic-apm-server:v1.2

  • The default secret_token have been changed from '' to xxVpmQB2HMzCL9PgBHVrnxjNXXw5J7bd79DFm6sjBJR5HPXDhcF8MSb3vv4bpg44

• with compose 2.2 feature, healthcheck

  • your may need to wait a while in the up stage
  • because kibana depends on es
  • apm-server depends on es and kibana

• apm-server may exit before elasticsearch is ready

  • exit status could be found by docker-compose ps
  • once exit then you should restart it, like docker-compose up -d

Resources

• docker elastic: https://www.docker.elastic.co/

• elasticsearch

  • image: docker pull docker.elastic.co/elasticsearch/elasticsearch:6.5.1
  • github: https://github.com/elastic/elasticsearch
  • docker: https://www.elastic.co/guide/en/elasticsearch/reference/6.0/docker.html
  • vm.max_map_count: docker-library/elasticsearch#111 (comment)

  $ grep vm.max_map_count /etc/sysctl.conf
  vm.max_map_count=262144
  
  $ sysctl -w vm.max_map_count=262144

• apm-server

  • image: docker pull docker.elastic.co/apm/apm-server:6.5.1
  • github: https://github.com/elastic/apm-server
  • docker: https://www.elastic.co/guide/en/apm/server/current/running-on-docker.html
  • config: https://github.com/elastic/apm-server/blob/master/apm-server.reference.yml

• kibana

  • image: docker pull docker.elastic.co/kibana/kibana:6.5.1
  • github: https://github.com/elastic/kibana
  • docker: https://www.elastic.co/guide/en/kibana/6.0/docker.html

• apm agent

  • nodejs: https://www.elastic.co/guide/en/apm/agent/nodejs/current/intro.html
    • agent api: https://www.elastic.co/guide/en/apm/agent/nodejs/current/agent-api.html
    • github: https://github.com/elastic/apm-agent-nodejs

Waterline

The higher the waterline, the higher the service load, formula is as follows

waterline = sum([waterline_0, waterline_1, ..., waterline_n]) / count(duration)

waterline_x = count(duration, [border_x_start, border_x_end]) * a_x

Where

• duration is transaction.duration.us
• border_x is interval border, like [0, 100] means from 0ms to 100ms
• a_x is interval coefficient of border_x, like 1/2/100 or whatever you want

Example

waterline = ( count(duration, [0, 200]) * 1 + count(duration, [200, *]) * 2 ) / count(duration)

In a point

10 = count(duration, [0, 200])
1 = count(duration, [200, *])

Then, in 3 decimal places precision

waterline = ( 10 * 1 + 1 * 2 ) / 11 = 1.091