Elasticsearch preload data to cache (qbit)

foreword

• This article applies to Elasticsearch 7.17
• Official documentation: https://www.elastic.co/guide/en/elasticsearch/reference/7.17/preload-data-to-file-system-cache.html

Which files to preload

• Official recommendation preload

   {
    "preload": [
        "nvd",
        "dvd",
        "tim",
        "doc",
        "dim"
    ]
  }

• There are many aggregations, but the memory is not very sufficient, you can consider preload

   {
    "preload": [
        "nvd",
        "dvd",
        "dvm"
    ]
  }

• If the content is sufficient, you can take the above union preload

   {
    "preload": [
        "nvd",
        "dvd",
        "dvm",
        "tim",
        "doc",
        "dim"
    ]
  }

other

• The official document says index.store.preload is a static configuration, which can be configured before elasticsearch.yml or before indexing. In fact, you can modify this configuration item after the index is established, change the configuration after the index close and then open .
• For the format and size of the index file, see: About ES File Format
• View index disk usage: Analyze index disk usage API

 # 在 kibana 上执行这条指令可能会超时
POST my-index/_disk_usage?run_expensive_tasks=true
# 可以换成 curl 执行指令
curl -X POST "localhost:9200/my-index/_disk_usage?run_expensive_tasks=true&pretty"
# 使用账号密码
curl --user elastic:my_pass -X POST "localhost:9200/my-index/_disk_usage?run_expensive_tasks=true&pretty"

• The official document does not mention that preloading can target specific fields, just use the field name: Engineering practice of Elasticsearch vector search
  

This article is from qbit snap