elasticsearch压力测试工具之ESrally使用说明

图片

大地
以行动承担责任,而非言辞

ESrally介绍

esrally是elastic search官方用于对ES集群进行压力测试的工具,使用esrally可以为我们构建不同版本集群,构造不同的参数和数据来进行压力测试,并且可以对产生的压测结果进行比较,rally顾名思义是拉力赛的意思,esrally的一些名词也都与拉力赛有关。
github地址:https://github.com/elastic/rally

ESrally安装

环境要求

python3.8 pip3
jdk8
git 1.9+

python3.8安装

yum install zlib-devel bzip2-devel openssl-devel ncurses-devel sqlite-devel readline-devel tk-devel gcc make libffi-devel 
wget https://www.python.org/ftp/python/3.8.2/Python-3.8.2.tar.xz
tar -xvJf Python-3.8.2.tar.xz
mkdir /usr/local/python3
cd Python-3.8.2/
./configure --prefix=/usr/local/python3
make && make install
ln -s /usr/local/python3/bin/python3 /usr/local/bin/python3
ln -s /usr/local/python3/bin/pip3 /usr/local/bin/pip3

git 2.22安装

由于yum安装的git版本默认为1.8,因此我需要编译安装较新版本的git,如果已通过yum安装过git,可以使用yum remove git卸载

yum install curl-devel expat-devel gettext-devel openssl-devel zlib-devel gcc perl-ExtUtils-MakeMaker
cd /tmp
wget https://mirrors.edge.kernel.org/pub/software/scm/git/git-2.22.0.tar.gz
tar xzf git-2.22.0.tar.gz
cd git-2.22.0
make prefix=/usr/local/git all
make prefix=/usr/local/git install
echo "export PATH=$PATH:/usr/local/git/bin" >> /etc/bashrc
source /etc/bashrc

?jdk安装

rpm -ivh jdk-8u221-linux-x64.rpm

?esrally安装

python3 -m pip install esrally
vim /etc/profile
JAVA_HOME=/usr/java/jdk1.8.0_221-amd64/jre
export PATH=$PATH:/usr/local/python3/bin/:/usr/local/git/bin JAVA_HOME
source /etc/profile

?配置ESrally

用于定义一些数据的配置,可以选择把结果存储到已有的ES中进行分析
esrally configure
配置完成后,将会覆写ESrally的配置文件/root/.rally/rally.ini
图片?ESrally名词解释

?track

赛道的意思,用于构建不同的数据和策略进行压测,track.json定义压测的数据集,以geonames/track.json为例

{% import "rally.helpers" as rally with context %}
{
  "version": 2,
  "description""POIs from Geonames",
  "data-url""http://benchmarks.elasticsearch.org.s3.amazonaws.com/corpora/geonames",
  "indices": [
    {
      "name""geonames",
      "body""index.json"
    }
  ],
  "corpora": [
    {
      "name""geonames",
      "base-url""http://benchmarks.elasticsearch.org.s3.amazonaws.com/corpora/geonames",
      "documents": [
        {
          "source-file""documents-2.json.bz2",
          "document-count": 11396503,
          "compressed-bytes": 265208777,
          "uncompressed-bytes": 3547613828
        }
      ]
    }
  ],
  "operations": [
    {{ rally.collect(parts="operations/*.json") }}
  ],
  "challenges": [
    {{ rally.collect(parts="challenges/*.json") }}
  ]
}

当我们开始用esrally开始比赛的时候,会先根据base-url和source-files构建链接从亚马逊下载数据,下载测试数据的速度很慢,我们可以提前下好离线数据来进行测试,测试数据存放在/home/elk/.rally/benchmarks/data/geonames目录下,indices定义了索引名和索引的具体设置,通过index.json文件可以查看

{
  "settings": {
    "index.number_of_shards": {{number_of_shards | default(5)}},
    "index.number_of_replicas": {{number_of_replicas | default(0)}},
    "index.store.type""{{store_type | default('fs')}}",
    "index.requests.cache.enable"false
  },
  "mappings": {
    "dynamic""strict",
    "_source": {
      "enabled": {{ source_enabled | default(true) | tojson }}
    },
    "properties": {
      "elevation": {
        "type""integer"
      },
      "name": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "geonameid": {
        "type""long"
      },
      "feature_class": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "location": {
        "type""geo_point"
      },
      "cc2": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "timezone": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "dem": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "country_code": {
        "type""text",
        "fielddata"true,
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "admin1_code": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "admin2_code": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "admin3_code": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "admin4_code": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "feature_code": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "alternatenames": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "asciiname": {
        "type""text",
        "fields": {
          "raw": {
            "type""keyword"
          }
        }
      },
      "population": {
        "type""long"
      }
    }
  }
}

operations目录下定义了压测的具体操作,包括索引的写入、更新、段合并、各种查询,每个操作都可以构建不同的参数,

{
      "name""index-append",
      "operation-type""bulk",
      "bulk-size": {{bulk_size | default(5000)}},
      "ingest-percentage": {{ingest_percentage | default(100)}}
    },
    {
      "name""index-update",
      "operation-type""bulk",
      "bulk-size": {{bulk_size | default(5000)}},
      "ingest-percentage": {{ingest_percentage | default(100)}},
      "conflicts""{{conflicts | default('random')}}",
      "on-conflict""{{on_conflict | default('index')}}",
      "conflict-probability": {{conflict_probability | default(25)}},
      "recency": {{recency | default(0)}}
    },
    {
      "name""default",
      "operation-type""search",
      "body": {
        "query": {
          "match_all": {}
        }
      }
    },
    {
      "name""term",
      "operation-type""search",
      "body": {
        "query": {
          "term": {
            "country_code.raw""AT"
          }
        }
      }
    },
    {
      "name""phrase",
      "operation-type""search",
      "body": {
        "query": {
          "match_phrase": {
            "name""Sankt Georgen"
          }
        }
      }
    },
    {
      "name""country_agg_uncached",
      "operation-type""search",
      "body": {
        "size": 0,
        "aggs": {
          "country_population": {
            "terms": {
              "field""country_code.raw"
            },
            "aggs": {
              "sum_population": {
                "sum": {
                  "field""population"
                }
              }
            }
          }
        }
      }
    },
    {
      "name""country_agg_cached",
      "operation-type""search",
      "cache"true,
      "body": {
        "size": 0,
        "aggs": {
          "country_population": {
            "terms": {
              "field""country_code.raw"
            },
            "aggs": {
              "sum_population": {
                "sum": {
                  "field""population"
                }
              }
            }
          }
        }
      }
    },
    {
      "name""scroll",
      "operation-type""search",
      "pages": 25,
      "results-per-page": 1000,
      "body": {
        "query": {
          "match_all": {}
        }
      }
    },
    {
      "name""expression",
      "operation-type""search",
      "body": {
        "query": {
          "function_score": {
            "query": {
              "match_all": {}
            },
            "functions": [
              {
                "script_score": {
                  "script": {
                    "source""abs(ln(abs(doc['population']) + 1) + doc['location'].lon + doc['location'].lat) * _score",
                    "lang""expression"
                  }
                }
              }
            ]
          }
        }
      }
    },
    {
      "name""painless_dynamic",
      "operation-type""search",
      "body": {
        "query": {
          "function_score": {
            "query": {
              "match_all": {}
            },
            "functions": [
              {
                "script_score": {
                  "script": {
                    "source""Math.abs(Math.log(Math.abs(doc['population'].value) + 1) + doc['location'].lon * doc['location'].lat)/_score",
                    "lang""painless"
                  }
                }
              }
            ]
          }
        }
      }
    },
    {
      "name""decay_geo_gauss_function_score",
      "operation-type""search",
      "body": {
        "query": {
          "function_score": {
            "query": {
              "match_all": {}
            },  
            "gauss": {
              "location": {
                "origin""52.37, 4.8951",
                "scale""500km",
                "offset""0km",
                "decay" : 0.1
              }
            }     
          }
        }
      }
    },
   {
      "name""decay_geo_gauss_script_score",
      "operation-type""search",
      "body": {
        "query": {
          "script_score": {
            "query": {
              "match_all": {}
            },
            "script": {
              "source""decayGeoGauss(params.origin, params.scale, params.offset, params.decay, doc['location'].value)",
              "params": {
                 "origin""52.37, 4.8951",
                  "scale""500km",
                  "offset""0km",
                  "decay" : 0.1 
              }
            }
          }
        }
      }
    },
    {
      "name""random_script_score",
      "operation-type""search",
      "body": {
        "query": {
          "script_score": {
            "query": {
              "match_all": {}
            },
            "script": {
              "source""randomScore(100, '_seq_no')"
            }
          }
        }
      }
    },     
    {
      "name""large_terms",
      "operation-type""search",
      "param-source""pure-terms-query-source"
    },
    {
      "name""desc_sort_population",
      "operation-type""search",
      "body": {
        "query": {
          "match_all": {}
        },
        "sort" : [
          {"population" : "desc"}
        ]
      }
    },
    {
      "name""asc_sort_population",
      "operation-type""search",
      "body": {
        "query": {
          "match_all": {}
        },
        "sort" : [
          {"population" : "asc"}
        ]
      }
    },
    {
      "name""desc_sort_geonameid",
      "operation-type""search",
      "body": {
        "query": {
          "match_all": {}
        },
        "sort" : [
          {"geonameid" : "desc"}
        ]
      }
    },
    {
      "name""asc_sort_geonameid",
      "operation-type""search",
      "body": {
        "query": {
          "match_all": {}
        },
        "sort" : [
          {"geonameid" : "asc"}
        ]
      }
    }

challenge指定了压测时执行的task,不同的track包含一个或多个challenge,"name": "append-no-conflicts"表示默认的challenge为append-no-conflicts,执行esrally list tracks可以查看不同track支持的challenge,default为true时,为默认执行的challenge,schedule定义了执行的任务列表

{
      "name""append-no-conflicts",
      "description""Indexes the whole document corpus using Elasticsearch default settings. We only adjust the number of replicas as we benchmark a single node cluster and Rally will only start the benchmark if the cluster turns green. Document ids are unique so all index operations are append only. After that a couple of queries are run.",
      "default"true,
      "schedule": [
        {
          "operation""delete-index"
        },
        {
          "operation": {
            "operation-type""create-index",
            "settings": {{index_settings | default({}) | tojson}}
          }
        },
        {
          "name""check-cluster-health",
          "operation": {
            "operation-type""cluster-health",
            "index""geonames",
            "request-params": {
              "wait_for_status""{{cluster_health | default('green')}}",
              "wait_for_no_relocating_shards""true"
            }
          }
        },
        {
          "operation""index-append",
          "warmup-time-period": 120,
          "clients": {{bulk_indexing_clients | default(8)}}
        },
        {
          "name""refresh-after-index",
          "operation""refresh"
        },
        {
          "operation": {
            "operation-type""force-merge",
            "request-timeout": 7200
          }
        },
        {
          "name""refresh-after-force-merge",
          "operation""refresh"
        },
        {
          "name""wait-until-merges-finish",
          "operation": {
            "operation-type""index-stats",
            "index""_all",
            "condition": {
              "path""_all.total.merges.current",
              "expected-value": 0
            },
            "retry-until-success"true,
            "include-in-reporting"false
          }
        },
        {
          "operation""index-stats",
          "warmup-iterations": 500,
          "iterations": 1000,
          "target-throughput": 90
        },
        {
          "operation""node-stats",
          "warmup-iterations": 100,
          "iterations": 1000,
          "target-throughput": 90
        },
        {
          "operation""default",
          "warmup-iterations": 500,
          "iterations": 1000,
          "target-throughput": 50
        },
        {
          "operation""term",
          "warmup-iterations": 500,
          "iterations": 1000,
          "target-throughput": 150
        },
        {
          "operation""country_agg_uncached",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 3.6
        },
        {
          "operation""country_agg_cached",
          "warmup-iterations": 1000,
          "iterations": 1000,
          "target-throughput": 100
        },
        {
          "operation""scroll",
          "warmup-iterations": 200,
          "iterations": 100,
          "#COMMENT""Throughput is considered per request. So we issue one scroll request per second which will retrieve 25 pages",
          "target-throughput": 0.8
        },
        {
          "operation""expression",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 2
        },
        {
          "operation""painless_static",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1.5
        },
        {
          "operation""painless_dynamic",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1.5
        },
        {
          "operation""decay_geo_gauss_function_score",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1
        },
        {
          "operation""decay_geo_gauss_script_score",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1
        },
        {
          "operation""field_value_function_score",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1.5
        },
        {
          "operation""field_value_script_score",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1.5
        },
        {
          "operation""random_function_score",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1.5
        },
        {
          "operation""large_filtered_terms",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1.1
        },
        {
          "operation""large_prohibited_terms",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 1.1
        },
        {
          "operation""desc_sort_geonameid",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 6
        },
        {
          "operation""asc_sort_geonameid",
          "warmup-iterations": 200,
          "iterations": 100,
          "target-throughput": 6
        }
      ]
    },
    {
      "name""append-fast-with-conflicts",
      "description""Indexes the whole document corpus using a setup that will lead to a larger indexing throughput than the default settings. Rally will produce duplicate ids in 25% of all documents (not configurable) so we can simulate a scenario with appends most of the time and some updates in between.",
      "schedule": [
        {
          "operation""delete-index"
        },
        {
          "operation": {
            "operation-type""create-index",
            "settings": {%- if index_settings is defined %} {{index_settings | tojson}} {%- else %} {
              "index.refresh_interval""30s",
              "index.number_of_shards": {{number_of_shards | default(6)}},
              "index.translog.flush_threshold_size""4g"
            }{%- endif %}
          }
        },
        {
          "name""check-cluster-health",
          "operation": {
            "operation-type""cluster-health",
            "index""geonames",
            "request-params": {
              "wait_for_status""{{cluster_health | default('green')}}",
              "wait_for_no_relocating_shards""true"
            }
          }
        },
        {
          "operation""index-update",
          "warmup-time-period": 45,
          "clients": {{bulk_indexing_clients | default(8)}}
        },
        {
          "operation": {
            "operation-type""force-merge",
            "request-timeout": 7200
          }
        },
        {
          "name""wait-until-merges-finish",
          "operation": {
            "operation-type""index-stats",
            "index""_all",
            "condition": {
              "path""_all.total.merges.current",
              "expected-value": 0
            },
            "retry-until-success"true,
            "include-in-reporting"false
          }
        }
      ]
    }

?car

用于定义不同配置的ES实例,我们可以定义堆内存大小,垃圾回收器,既然是赛车,那就可以改装,我们可以自定义不同配置的car,执行命令esrally list car可查看所有的赛车,配置在/home/elk/.rally/benchmarks/teams/default/cars/v1目录下
图片?race

表示某一次的压测,并且可以指定赛道和赛车配置,不指定则使用默认配置,race结果存储在/home/elk/.rally/benchmarks/races目录下,执行命令esrally list races可查看以往比赛结果
图片并且可以对不同的race进行比较

esrally compare --baseline=d1be27e5-c459-4a40-ab7b-f3576cda7144 --contender=694f5c89-2ed1-4a82-8524-98a8e59778d4

图片?pipeline

就是定义以何种方式来构建这个集群,通过esrally list pipelines可查看所有的pipeline
图片

  • from-sources-complete :表示从源代码构建ES, --revision参数可以指定一个ES的修订版本,默认为最新版本、
  • from-sources-skip-build :表示如果之前已经通过源码构建过一次相同版本的ES了,那么本次将不再进行构建,这样可以节省测试的时间
  • from-distribution:esrally将下载正式的ES发行版来进行测试, --distribution-version 参数可以指定ES版本,最低支持1.7.5
  • benchmark-only :将对已有的ES集群进行压测,--target-hosts参数可以指定ES集群地址

ESrally压测实例

?测试5.4.3、6.4.3与7.8.1版本之间的写入性能差异

使用非root用户运行,并且保证内存够用,这里我是用geonames赛道进行测试,使用默认的car为1gb堆内存,为了快速测试,可以将测试数据和ES二进制包提前下到指定的目录下,ES二进制包存储在/home/elk/.rally/benchmarks/distributions目录下,测试数据存放在/home/elk/.rally/benchmarks/data/geonames目录下

esrally race --distribution-version=5.4.3 --track=geonames  --user-tag="version:5.4.3" --include-tasks="type:bulk"
esrally race --distribution-version=6.4.3 --track=geonames  --user-tag="version:6.4.3" --include-tasks="type:bulk"
esrally race --distribution-version=7.8.1 --track=geonames  --user-tag="version:7.8.1"  --include-tasks="type:bulk"

压测过程可以通过/home/elk/.rally/logs.log查看
图片?对压测结果进行比较

比较7.8.1与5.4.3的写入差异

esrally compare --baseline=27265e6e-566a-4a47-a0d9-1fd2f8830041 --contender=66086ef0-5834-4743-a870-fd9c0bb41688

图片可以看到5.4版本的写入性能差别还是很大的
图片再看一下6.4.3与7.8.1的差异,写入性能差别不是很大

esrally compare --baseline=27265e6e-566a-4a47-a0d9-1fd2f8830041 --contender=b02fb6fe-824f-48ee-beb5-ab2ca2ad4bbc

图片?测试x86_64平台与arm平台的性能差异

?环境

阿里云2v4g  志强Platinum 8269CY CPU 2.50GHz  x86_64
华为云2v4g  鲲鹏920  CPU 2.6GHz    arm平台
两台云主机的ES配置均相同,2G堆内存,系统参数相同

开始测试

还是要提前下好测试数据,这个测试周日跑了一天,结果截图没保存,可能配置太低了,结果差别不是很大,而且不是在同一台机器上跑的,没法使用esrally compare进行比较

esrally race --pipeline=benchmark-only --target-hosts=172.16.0.95:9200 --track=http_logs --offline
esrally race --pipeline=benchmark-only --target-hosts=172.26.214.32:9200 --track=http_logs --offline

?中间遇到的问题

开始压测之后,报不能进入比赛,pid文件不可用
图片之后查看了相关race的日志,发现内存不够用了,因为之前已经跑了一个ES实例了,导致了我内存不足,把之前的ES实例关掉即可解决
图片考链接:https://esrally.readthedocs.io/en/stable/quickstart.html

- END -

公众号:运维开发故事

github:https://github.com/orgs/sunsharing-note/dashboard

爱生活,爱运维

如果你觉得文章还不错,就请点击右上角选择发送给朋友或者转发到朋友圈。您的支持和鼓励是我最大的动力。喜欢就请关注我吧~

图片

扫码二维码

关注我,不定期维护优质内容

温馨提示

如果你喜欢本文,请分享到朋友圈,想要获得更多信息,请关注我。

                                              ........................

------本页内容已结束,喜欢请分享------

© 版权声明
THE END
喜欢就支持一下吧
点赞0
分享