Elasticsearch InOut Plugin

This Elasticsearch plugin provides the ability to export data by query on server side, by outputting the data directly on the according node. The export can happen on all indexes, on a specific index or on a specific document type.

The data will get exported as one json object per line:

{"_id":"id1","_source":{"type":"myObject","value":"value1"},"_version":1,"_index":"myIndex","_type":"myType"}
{"_id":"id2","_source":{"type":"myObject","value":"value2"},"_version":2,"_index":"myIndex","_type":"myType"}

The exported data can also be imported into elasticsearch again. The import is happening on each elasticsearch node by processing the files located in the specified directory.

Examples

Below are some examples demonstrating what can be done with the elasticsearch inout plugin. The example commands require installation on a UNIX system. The plugin may also works with different commands on other operating systems supporting elasticsearch, but is not tested yet.

Export data to files in the node's file system. The filenames will be expanded by index and shard names (p.e. /tmp/dump-myIndex-0):

curl -X POST 'http://localhost:9200/_export' -d '{
    "fields": ["_id", "_source", "_version", "_index", "_type"],
    "output_file": "/tmp/es-data/dump-${index}-${shard}"
}
'

Do GZIP compression on file exports:

curl -X POST 'http://localhost:9200/_export' -d '{
    "fields": ["_id", "_source", "_version", "_index", "_type"],
    "output_file": "/tmp/es-data/dump-${index}-${shard}.gz",
    "compression": "gzip"
}
'

Pipe the export data through a single argumentless command on the corresponding node, like cat. This command actually returns the export data in the JSON result's stdout field:

curl -X POST 'http://localhost:9200/_export' -d '{
    "fields": ["_id", "_source", "_version", "_index", "_type"],
    "output_cmd": "cat"
}
'

Pipe the export data through argumented commands (p.e. a shell script, or provide your own sophisticated script on the node). This command will result in transforming the data to lower case and write the file to the node's file system:

curl -X POST 'http://localhost:9200/_export' -d '{
    "fields": ["_id", "_source", "_version", "_index", "_type"],
    "output_cmd": ["/bin/sh", "-c", "tr [A-Z] [a-z] > /tmp/outputcommand.txt"]
}
'

Limit the exported data with a query. The same query syntax as for search can be used:

curl -X POST 'http://localhost:9200/_export' -d '{
    "fields": ["_id", "_source", "_version", "_index", "_type"],
    "output_file": "/tmp/es-data/query-${index}-${shard}",
    "query": {
        "match": {
            "someField": "someValue"
        }
    }
}
'

Export only objects of a specifix index:

curl -X POST 'http://localhost:9200/myIndex/_export' -d '{
    "fields": ["_id", "_source", "_version", "_type"],
    "output_file": "/tmp/es-data/dump-${index}-${shard}"
}
'

Export only objects of a specific type of an index:

curl -X POST 'http://localhost:9200/myIndex/myType/_export' -d '{
    "fields": ["_id", "_source", "_version"],
    "output_file": "/tmp/es-data/dump-${index}-${shard}"
}
'

Import data from previously exported data into elastic search. This can be for example a new set up elasticsearch server with empty indexes. Take care to have the indexes prepared with correct mappings. The files must reside on the file system of the elastic search node(s):

curl -X POST 'http://localhost:9200/_import' -d '{
    "directory": "/tmp/es-data"
}
'

Import data of gzipped files:

curl -X POST 'http://localhost:9200/_import' -d '{
    "directory": "/tmp/es-data",
    "compression": "gzip"
}
'

Import data into a specific index. Can be used if no _index is given in the export data or to force data of other indexes to be imported into a specific index:

curl -X POST 'http://localhost:9200/myNewIndex/_import' -d '{
    "directory": "/tmp/es-data"
}
'

Import data into a specific type of an index:

curl -X POST 'http://localhost:9200/myNewIndex/myType/_import' -d '{
    "directory": "/tmp/es-data"
}
'

Use a regular expression to filter imported file names (e.g. for specific indexes):

curl -X POST 'http://localhost:9200/_import' -d '{
    "directory": "/tmp/es-data",
    "file_pattern": "dump-myindex-(\\d).json"
}
'

Exports

Elements of the request body

fields

A list of fields to export. Describes which data is exported for every object. A field name can be any property that is defined in the index/type mapping with "store": "yes" or one of the following special fields (prefixed with _):

• _id: Delivers the ID of the object
• _index: Delivers the index of the object
• _routing: Delivers the routing value of the object
• _source: Delivers the stored JSON values of the object
• _timestamp: Delivers the time stamp when the object was created (or the externally provided timestamp). Works only if the _timestamp field is enabled and set to "store": "yes" in the index/type mapping of the object.
• _ttl: Delivers the expiration time stamp of the object if the _ttl field is enabled in the index/type mapping.
• _type: Delivers the document type of the object
• _version: Delivers the current version of the object

Example assuming that the properties name and address are defined in the index/type mapping with the property "store": "yes":

"fields": ["_id", "name", "address"]

The fields element is required in the POST data of the request.

output_cmd

"output_cmd": "cat"

"output_cmd": ["/location/yourcommand", "argument1", "argument2"]

The command to execute. Might be defined as string or as array. The content to export will get piped to Stdin of the command to execute. Some variable substitution is possible (see Variable Substitution)

• Required (if output_file has been omitted)

output_file

"output_file": "/tmp/dump"

A path to the resulting output file. The containing directory of the given output_file has to exist. The given output_file MUST NOT exist, unless the parameter force_overwrite is set to true.

If the path of the output file is relative, the files will be stored relative to each node's first node data location, which is usually a subdirectory of the configured data location. This absolute path can be seen in the JSON response of the request. If you don't know where this location is, you can do a dry-run with the explain element set to true to find out.

Some variable substitution in the output_file's name is also possible (see Variable Substitution).

• Required (if output_cmd has been omitted)

force_overwrite

"force_overwrite": true

Boolean flag to force overwriting existing output_file. This option only make sense if output_file has been defined.

• Optional (defaults to false)

explain

"explain": true

Option to evaluate the command to execute (like dry-run).

• Optional (defaults to false)

compression

"compression": "gzip"

Option to activate compression to the output. Works both whether output_file or output_cmd has been defined. Currently only the gzip compression type is available. Omitting the option will result in uncompressed output to files or processes.

• Optional (default is no compression)

query

The query element within the export request body allows to define a query using the Query DSL. See http://www.elasticsearch.org/guide/reference/query-dsl/

• Optional

settings

"settings": true

Option to generate an index settings file next to the data files on all corresponding shards. The generated settings file has the generated name of the output file with the .settings extension. This option is only possible if the option output_file has been defined.

• Optional (defaults to false)

mappings

"mappings": true

Option to generate an index mapping file next to the data files on all corresponding shards. The generated mapping file has the generated name of the output file with an .mapping extension. This option is only possible if the option output_file has been defined.

• Optional (defaults to false)

Get parameters

The api provides the general behavior of the rest API. See http://www.elasticsearch.org/guide/reference/api/

Preference

Controls a preference of which shard replicas to execute the export request on. Different than in the search API, preference is set to "_primary" by default. See http://www.elasticsearch.org/guide/reference/api/search/preference/

Variable Substitution

The following placeholders will be replaced with the actual value in the output_file or output_cmd fields:

• ${cluster}: The name of the cluster
• ${index}: The name of the index
• ${shard}: The id of the shard

JSON Response

The _export query returns a JSON response with information about the export status. The output differs a bit whether an output command or an output file is given in the request body.

Output file JSON response

The JSON response may look like this if an output file is given in the request body:

{
    "exports" : [
        {
            "index" : "myIndex",
            "shard" : 0,
            "node_id" : "the_node_id",
            "numExported" : 5,
            "output_file" : "/tmp/dump-myIndex-0"
        }
    ],
    "totalExported" : 5,
    "_shards" : {
        "total" : 2,
        "successful" : 1,
        "failed" : 1,
        "failures" : [
            {
                "index" : "myIndex",
                "shard" : 1,
                "reason" : "..."
            }
        ]
    }
}

Output command JSON response

The JSON response may look like this if an output command is given in the request body:

{
    "exports" : [
        {
            "index" : "myIndex",
            "shard" : 0,
            "node_id" : "the_node_id",
            "numExported" : 5,
            "output_cmd" : [
                "/bin/sh",
                "-c",
                "tr [A-Z] [a-z] > /tmp/outputcommand.txt"
            ],
            "stderr" : "",
            "stdout" : "",
            "exitcode" : 0
        }
    ],
    "totalExported" : 5,
    "_shards" : {
        "total" : 2,
        "successful" : 1,
        "failed" : 1,
        "failures": [
            {
                "index" : "myIndex",
                "shard" : 1,
                "reason" : "..."
            }
        ]
    }
}

Imports

Import data

The import data requires the same format as the format that is generated by the export. So every line in the import file represents an object in JSON format.

The _source field is required for a successful import of an object. If the _id field is not given, a random id is generated for the object. Also the _index and _type fields are required, as long as they are not given in the request URI (p.e. http://localhost:9200/<index>/<type>/_index).

Further optional fields are _routing, _timestamp, _ttl and _version. See the fields section on export for more details on the fields.

Elements of the request body

directory

Specifies the directory where the files to be imported reside. Every single node of the cluster imports files from that directory on it's file system.

If the directory is a relative path, it is based on the absolute path of each node's first node data location. See output_file in export documentation for more information.

compression

"compression": "gzip"

Option to activate decompression on the import files. Currently only the gzip compression type is available.

• Optional (default is no decompression)

file_pattern

"file_pattern": "index-(.*)-(\d).json"

Option to import only files with a given regular expression. Take care of double escaping, as the JSON is decoded too in the process. For more information on regular expressions visit http://www.regular-expressions.info/

• Optional (default is no filtering)

settings

"settings": true

Option to import index settings. All files in the import directory with an eponymic data file without the .settings extension will be handled. Also use the file_pattern option to reduce imported settings files. The format of a settings file is the same as the JSON output of _settings GET requests.

• Optional (defaults to false)

mappings

"mappings": true

Option to import index mappings. All files in the import directory with an eponymic data file without the .mapping extension will be handled. Also use the file_pattern option to reduce imported mapping files. The format of a mapping file is the same as the JSON output of _mapping GET requests.

• Optional (defaults to false)

JSON Response

The JSON response of an import may look like this:

{
    "imports" : [
        {
            "node_id" : "7RKUKxNDQlq0OzeOuZ02pg",
            "took" : 61,
            "imported_files" : [
                {
                    "file_name" : "dump-myIndex-1.json",
                    "successes" : 150,
                    "failures" : 0
                },
                {
                    "file_name" : "dump-myIndex-2.json",
                    "successes" : 149,
                    "failures" : 1,
                    "invalidated" : 1
                }
            ]
        },
        {
            "node_id" : "IrMCOlKCTtW4aDhjXiYzTw",
            "took" : 63,
            "imported_files" : [
                {
                    "file_name" : "dump-myIndex-3.json",
                    "successes" : 150,
                    "failures" : 0
                }
            ]
        }
    ],
    "failures" : [
        {
            "node_id" : "OATwHz48TEOshAISZlepcA",
            "reason" : "..."
        }
    ]
}

Dump

The idea behind dump is to export all relevant data to recreate the cluster as it was at the time of the dump.

The basic usage of the endpoint is:

curl -X POST 'http://localhost:9200/_dump'

All data (including also settings and mappings) will get saved to a subfolder within each nodes data directory.

It's possible to call _dump on root level, on index level or on type level.

Elements of the request body

directory

The directory option defines where to store exported files. If the directory is a relative path, it is based on the absolute path of each node's first node data location. See output_file in export documentation for more information. If the directory was omitted the default location dump within the node data location will be used.

force_overwrite

"force_overwrite": true

Boolean flag to force overwriting existing output_file. This option is identical to the force_overwrite option of the _export endpoint.

Restore

Dumped data is intended to get restored. This can be done by the _restore endpoint:

curl -X POST 'http://localhost:9200/_restore'

It's possible to call _restore on root level, on index level or on type level.

Elements of the request body

directory

Specifies the directory where the files to be restored reside. See directory in import documentation for more details. If the directory was omitted the default location dump within the node data location will be used.

settings and mappings

Defaults to true on restore. See the Import documentation for more details.

Reindex

The _reindex endpoint can reindex documents of a given search query.

Reindex all indexes:

curl -X POST 'http://localhost:9200/_reindex'

Reindex a specific index:

curl -X POST 'http://localhost:9200/myIndex/_reindex'

Reindex documents of a specified query:

curl -X POST 'http://localhost:9200/myIndex/aType/_reindex' -d '{
    "query": {"text": {"name": "tobereindexed"}}
}'

An example can be found in the Reindex DocTest.

Search Into

Via the _search_into endpoint it is possible to put the result of a given query directly into an index:

curl -X POST 'http://localhost:9200/oldindex/_search_into -d '{
    "fields": ["_id", "_source", ["_index", "'newindex'"]]
}'

An example can be found in the Search Into DocTest.

Installation

• Clone this repo with git clone [email protected]:crate/elasticsearch-inout-plugin.git
• Checkout the tag (find out via git tag) you want to build with (possibly master is not for your elasticsearch version)
• Run: mvn clean package -DskipTests=true – this does not run any unit tests, as they take some time. If you want to run them, better run mvn clean package
• Install the plugin: /path/to/elasticsearch/bin/plugin -install elasticsearch-inout-plugin -url file:///$PWD/target/elasticsearch-inout-plugin-$version.jar