Reducing requested filter cache size of to the maximum allowed size of – How to solve this Elasticsearch error

Overview

Elasticsearch uses three types of caches to improve the efficiency of operation.  

• Node request cache
• Shard data cache
• Field data cache

How they work

Node request cache maintains the results of queries used in a filter context. The results are evicted on a least recently used basis.

Shard data cache maintains the results of frequently used queries where size=0, particularly the results of aggregations.  This cache is particularly relevant for logging use cases where data is not updated on old indices, and regular aggregations can be kept in cache to be reused.

The field data cache is used for sorting and aggregations. To keep these operations quick Elasticsearch loads these values into memory.   

Examples

Elasticsearch usually manages cache behind the scenes, without the need for any specific settings. However, it is possible to monitor and limit the amount of memory being used on each node for a given cache type by putting the following in elasticsearch.yml :

indices.queries.cache.size: 10%

indices.fielddata.cache.size: 30%

Note, the above values are in fact the defaults, and there is no need to set them specifically. The default values are good for most use cases, and should rarely be modified.
You can monitor the use of caches on each node like this:

GET /_nodes/stats/indices/fielddata

GET /_nodes/stats/indices/query_cache

GET /_nodes/stats/indices/request_cache

Notes and good things to know

Construct your queries with reusable filters. There are certain parts of your query which are good candidates to be reused across a large number of queries, and you should design your queries with this in mind. Anything thing that does not need to be scored should go in the filter section of a bool query. For example, time ranges, language selectors, or clauses that exclude inactive documents are all likely to be excluded in a large number of queries, and should be included in filter parts of the query so that they can be cached and reused. 

In particular, take care with time filters. “now-15m” cannot be reused, because “now” will continually change as the time window moves on. On the other hand “now-15/m” will round to the nearest minute, and can be re-used (via cache) for 60 seconds before rolling over to the next minute.

For example when a user enters the search term “brexit”, we may want to also filter on language and time period to return relevant articles. The query below leaves only the query term “brexit” in the “must” part of the query, because this is the only part which should affect the relevance score. The time filter and language filter can be reused time and time again for new queries for different searches.

POST results/_search
{
  "query": {
	"bool": {
  	"must": [
    	{
      	"match": {
        	"message": {
          	"query": "brexit"
        	}
      	}
    	}
  	],
  	"filter": [
    	{
      	"range": {
        	"@timestamp": {
          	"gte": "now-10d/d"
          	        	}
      	}
    	},
    	{
      	"term": {
        	"lang.keyword": {
          	"value": "en",
          	"boost": 1
        	}
      	}
    	}
  	]
	}
  }
}

Limit the use of field data. Be careful about using fielddata=true in your mapping where the number of terms will result in a high cardinality.  If you must use fielddata=true, you can also reduce the requirement of fielddata cache by limiting the requirements for fielddata for a given index using a field data frequency filter.

POST results/_search
{
  "query": {
	"bool": {
  	"must": [
    	{
      	"match": {
        	"message": {
          	"query": "brexit"
        	}
      	}
    	}
  	],
  	"filter": [
    	{
      	"range": {
        	"@timestamp": {
          	"gte": "now-10d/d"
          	        	}
      	}
    	},
    	{
      	"term": {
        	"lang.keyword": {
          	"value": "en",
          	"boost": 1
        	}
      	}
    	}
  	]
	}
  }
}

Overview

When a query is executed, Elasticsearch by default calculates the relevance score of the matching documents. But in some conditions, it does not require scores to be calculated. For instance, if a document falls in the range of two given timestamps or if a document contains a given list of tags. For all these Yes/No criteria, also known as structured search, a filter clause is used.

When it is not desired or not necessary to compute scores, filters should be used instead of queries, as frequently used filters can be cached automatically by Elasticsearch to improve performance.

There are multiple ways to specify filters, such as when using the `filter` and `must_not` parameters of the `bool` query, the `filter` parameter of the `constant_score` query or the `filter` aggregation.

What it is used for

When a query is executed, Elasticsearch by default calculates the relevance score of the matching documents. But in some conditions it does not require scores to be calculated, for instance if a document falls in the range of two given timestamps. For all these Yes/No criteria, a filter clause is used.

Examples

To return all the documents of a given index that fall between a date range, we can use the `range` filter, as shown below:

GET my_index/_search
{
  "query": {
    "bool": {
      "filter": [
        {
          "range": {
            "created_at": {
              "gte": "2020-01-01",
              "lte": "2020-01-10"
            }
          }
        }
      ]
    }
  }
}

To retrieve all the documents that contain at least one tag from a given list, we can use the `terms` filter, as shown below:

GET my_index/_search
{
  "query": {
    "bool": {
      "filter": [
        {
          "terms": {
            "tags": ["tag1", "tag2", "tag3"]
          }
        }
      ]
    }
  }
}

To retrieve all the documents that contain a given field having a non-null value, we can use the `exists` filter, as shown below:

GET my_index/_search
{
  "query": {
    "bool": {
      "filter": [
        {
          "exists": {
            "field": "field_name"
          }
        }
      ]
    }
  }
}

There are many other filters that we can use in order to reduce the document set that needs to be scored, such as `fuzzy`, `prefix`, `wildcard`, `regexp`, `script`, and many more. 

It is also worth noting that filters can be combined since the `bool/filter` and `bool/must_not` parameters are arrays. In the example below, we retrieve all documents falling within a data range, containing a list of tags and not having a specific field:

GET my_index/_search
{
  "query": {
    "bool": {
      "filter": [
        {
          "range": {
            "created_at": {
              "gte": "2020-01-01",
              "lte": "2020-01-10"
            }
          }
        },
        {
          "terms": {
            "tags": ["tag1", "tag2", "tag3"]
          }
        }
      ],
      "must_not": [
        {
          "exists": {
            "field": "field_name"
          }
        }
      ]
    }
  }
}

Notes

• Queries are used to find out how relevant a document is to a particular query by calculating a score for each document, whereas filters are used to match certain criteria and are cacheable to enable faster execution.
• Filters do not contribute to scoring and thus are faster to execute.
• There are major changes introduced in Elasticsearch version 2.x onward related to how query and filters are written and performed internally and each newer version comes with its load of new improvements.

Common problems

• The most common problem with filters is incorrect use inside the query. If filters are not used correctly, query performance can be significantly affected. So filters must be used wherever there is scope of not calculating the score. 
• Another problem often arises when using date range filters, if “now” is used to represent the current time. It has to be noted that “now” is continuously changing the timestamp and thus Elasticsearch cannot use caching of the response since the data set will keep changing.

Overview

In Elasticsearch, an index (plural: indices) contains a schema and can have one or more shards and replicas. An Elasticsearch index is divided into shards and each shard is an instance of a Lucene index.

Indices are used to store the documents in dedicated data structures corresponding to the data type of fields. For example, text fields are stored inside an inverted index whereas numeric and geo fields are stored inside BKD trees.

Examples

Create index

The following example is based on Elasticsearch version 5.x onwards. An index with two shards, each having one replica will be created with the name test_index1

PUT /test_index1?pretty
{
    "settings" : {
        "number_of_shards" : 2,
        "number_of_replicas" : 1
    },
    "mappings" : {
        "properties" : {
            "tags" : { "type" : "keyword" },
            "updated_at" : { "type" : "date" }
        }
    }
}

List indices

All the index names and their basic information can be retrieved using the following command:

GET _cat/indices?v

Index a document

Let’s add a document in the index with the command below:

PUT test_index1/_doc/1
{
  "tags": [
    "opster",
    "elasticsearch"
  ],
  "date": "01-01-2020"
}

Query an index

GET test_index1/_search
{
  "query": {
    "match_all": {}
  }
}

Query multiple indices

It is possible to search multiple indices with a single request. If it is a raw HTTP request, index names should be sent in comma-separated format, as shown in the example below, and in the case of a query via a programming language client such as python or Java, index names are to be sent in a list format.

GET test_index1,test_index2/_search

Delete indices

DELETE test_index1

Common problems

DELETE /*

action.destructive_requires_name: true