search_into.rst

Search Into

Via the _search_into endpoint it is possible to put the result of a given query directly into an index. It is inspired from the Select Into SQL-Statement

Here are some common use-cases that might be fulfilled by using this endpoint.

Copy an existing index

To copy a whole index, the _index field source needs to be defined as a string literal like this:

>>> fields = ["_id", "_source", ["_index", "'newindex'"]]

There are two special things happening here. When defining a field as an array of 2 items, the first item is considered as the target-field name and the second item is considered as the source value. The above example defines a string literal (denoted by the surrounding apostrophes) as the value of the _index field.

>>> post("/users/_search_into?pretty=1", {"fields":fields})
{
  "writes" : [ {
    "index" : "users",
    "shard" : 0,
    "node" : "...",
    "total" : 2,
    "succeeded" : 2,
    "failed" : 0
  }, {
    "index" : "users",
    "shard" : 1,
    "node" : "...",
    "total" : 2,
    "succeeded" : 2,
    "failed" : 0
  }...

Now a new index with the same content was created:

>>> post("/newindex/_flush")
{"ok":true,"_shards":{"total":1,"successful":1,"failed":0}}

>>> post("/newindex/_refresh")
{"ok":true,"_shards":{"total":1,"successful":1,"failed":0}}


>>> get("/newindex/_search?pretty=1")
{..."hits" : {
    "total" : 4,
    ...{
      "_index" : "newindex",
      "_type" : "d",
      "_id" : "1",
      "_score" : 1.0, "_source" : {"name":"car"}
    }...

Copy an existing index into another cluster

To copy one index to another cluster set the option targetNodes in the payload. The value of the option might be given as a single value in the format <hostname>:<port> or as a list of such values.

Initially, the second cluster does not contain any indizes:

>>> node2.get("/_status?pretty=1")
{
  ...
  "indices" : { }
}

>>> payload = {"fields":["_id", "_source"],
...            "targetNodes":"localhost:9301"}

>>> post("/users/_search_into?pretty=1", payload)
{
  "writes" : [ {
    "index" : "users",
    "shard" : 0,
    "node" : "...",
    "total" : 2,
    "succeeded" : 2,
    "failed" : 0
  }, {
    "index" : "users",
    "shard" : 1,
    "node" : "...",
    "total" : 2,
    "succeeded" : 2,
    "failed" : 0
  } ],
  "total" : 0,
  "succeeded" : 0,
  "failed" : 0,
  "_shards" : {
    "total" : 2,
    "successful" : 2,
    "failed" : 0
  }
}

Now the new index with the same content was created in the other cluster:

>>> node2.post("/users/_flush")
{"ok":true,"_shards":{"total":1,"successful":1,"failed":0}}

>>> node2.post("/users/_refresh")
{"ok":true,"_shards":{"total":1,"successful":1,"failed":0}}

>>> node2.get("/users/_search?pretty=1")
{..."hits" : {
    "total" : 4,
    ...{
      "_index" : "users",
      "_type" : "d",
      "_id" : "1",
      "_score" : 1.0, "_source" : {"name":"car"}
    }...