[[pipeline]] == Pipeline Definition A pipeline is a definition of a series of <> that are to be executed in the same order as they are declared. A pipeline consists of two main fields: a `description` and a list of `processors`: [source,js] -------------------------------------------------- { "description" : "...", "processors" : [ ... ] } -------------------------------------------------- // NOTCONSOLE The `description` is a special field to store a helpful description of what the pipeline does. The `processors` parameter defines a list of processors to be executed in order. [[ingest-apis]] == Ingest APIs The following ingest APIs are available for managing pipelines: * <> to add or update a pipeline * <> to return a specific pipeline * <> to delete a pipeline * <> to simulate a call to a pipeline [[put-pipeline-api]] === Put Pipeline API The put pipeline API adds pipelines and updates existing pipelines in the cluster. [source,js] -------------------------------------------------- PUT _ingest/pipeline/my-pipeline-id { "description" : "describe pipeline", "processors" : [ { "set" : { "field": "foo", "value": "bar" } } ] } -------------------------------------------------- // CONSOLE NOTE: The put pipeline API also instructs all ingest nodes to reload their in-memory representation of pipelines, so that pipeline changes take effect immediately. [[get-pipeline-api]] === Get Pipeline API The get pipeline API returns pipelines based on ID. This API always returns a local reference of the pipeline. [source,js] -------------------------------------------------- GET _ingest/pipeline/my-pipeline-id -------------------------------------------------- // CONSOLE // TEST[continued] Example response: [source,js] -------------------------------------------------- { "my-pipeline-id" : { "description" : "describe pipeline", "processors" : [ { "set" : { "field" : "foo", "value" : "bar" } } ] } } -------------------------------------------------- // TESTRESPONSE For each returned pipeline, the source and the version are returned. The version is useful for knowing which version of the pipeline the node has. You can specify multiple IDs to return more than one pipeline. Wildcards are also supported. [float] [[versioning-pipelines]] ==== Pipeline Versioning Pipelines can optionally add a `version` number, which can be any integer value, in order to simplify pipeline management by external systems. The `version` field is completely optional and it is meant solely for external management of pipelines. To unset a `version`, simply replace the pipeline without specifying one. [source,js] -------------------------------------------------- PUT _ingest/pipeline/my-pipeline-id { "description" : "describe pipeline", "version" : 123, "processors" : [ { "set" : { "field": "foo", "value": "bar" } } ] } -------------------------------------------------- // CONSOLE To check for the `version`, you can <> using `filter_path` to limit the response to just the `version`: [source,js] -------------------------------------------------- GET /_ingest/pipeline/my-pipeline-id?filter_path=*.version -------------------------------------------------- // CONSOLE // TEST[continued] This should give a small response that makes it both easy and inexpensive to parse: [source,js] -------------------------------------------------- { "my-pipeline-id" : { "version" : 123 } } -------------------------------------------------- // TESTRESPONSE [[delete-pipeline-api]] === Delete Pipeline API The delete pipeline API deletes pipelines by ID or wildcard match (`my-*`, `*`). [source,js] -------------------------------------------------- DELETE _ingest/pipeline/my-pipeline-id -------------------------------------------------- // CONSOLE // TEST[continued] //// Hidden setup for wildcard test: [source,js] -------------------------------------------------- PUT _ingest/pipeline/wild-one { "description" : "first pipeline to be wildcard deleted", "processors" : [ ] } PUT _ingest/pipeline/wild-two { "description" : "second pipeline to be wildcard deleted", "processors" : [ ] } DELETE _ingest/pipeline/* -------------------------------------------------- // CONSOLE Hidden expected response: [source,js] -------------------------------------------------- { "acknowledged": true } -------------------------------------------------- // TESTRESPONSE //// [[simulate-pipeline-api]] === Simulate Pipeline API The simulate pipeline API executes a specific pipeline against the set of documents provided in the body of the request. You can either specify an existing pipeline to execute against the provided documents, or supply a pipeline definition in the body of the request. Here is the structure of a simulate request with a pipeline definition provided in the body of the request: [source,js] -------------------------------------------------- POST _ingest/pipeline/_simulate { "pipeline" : { // pipeline definition here }, "docs" : [ { "_source": {/** first document **/} }, { "_source": {/** second document **/} }, // ... ] } -------------------------------------------------- // NOTCONSOLE Here is the structure of a simulate request against an existing pipeline: [source,js] -------------------------------------------------- POST _ingest/pipeline/my-pipeline-id/_simulate { "docs" : [ { "_source": {/** first document **/} }, { "_source": {/** second document **/} }, // ... ] } -------------------------------------------------- // NOTCONSOLE Here is an example of a simulate request with a pipeline defined in the request and its response: [source,js] -------------------------------------------------- POST _ingest/pipeline/_simulate { "pipeline" : { "description": "_description", "processors": [ { "set" : { "field" : "field2", "value" : "_value" } } ] }, "docs": [ { "_index": "index", "_type": "_doc", "_id": "id", "_source": { "foo": "bar" } }, { "_index": "index", "_type": "_doc", "_id": "id", "_source": { "foo": "rab" } } ] } -------------------------------------------------- // CONSOLE Response: [source,js] -------------------------------------------------- { "docs": [ { "doc": { "_id": "id", "_index": "index", "_type": "_doc", "_source": { "field2": "_value", "foo": "bar" }, "_ingest": { "timestamp": "2017-05-04T22:30:03.187Z" } } }, { "doc": { "_id": "id", "_index": "index", "_type": "_doc", "_source": { "field2": "_value", "foo": "rab" }, "_ingest": { "timestamp": "2017-05-04T22:30:03.188Z" } } } ] } -------------------------------------------------- // TESTRESPONSE[s/"2017-05-04T22:30:03.187Z"/$body.docs.0.doc._ingest.timestamp/] // TESTRESPONSE[s/"2017-05-04T22:30:03.188Z"/$body.docs.1.doc._ingest.timestamp/] [[ingest-verbose-param]] ==== Viewing Verbose Results You can use the simulate pipeline API to see how each processor affects the ingest document as it passes through the pipeline. To see the intermediate results of each processor in the simulate request, you can add the `verbose` parameter to the request. Here is an example of a verbose request and its response: [source,js] -------------------------------------------------- POST _ingest/pipeline/_simulate?verbose { "pipeline" : { "description": "_description", "processors": [ { "set" : { "field" : "field2", "value" : "_value2" } }, { "set" : { "field" : "field3", "value" : "_value3" } } ] }, "docs": [ { "_index": "index", "_type": "_doc", "_id": "id", "_source": { "foo": "bar" } }, { "_index": "index", "_type": "_doc", "_id": "id", "_source": { "foo": "rab" } } ] } -------------------------------------------------- // CONSOLE Response: [source,js] -------------------------------------------------- { "docs": [ { "processor_results": [ { "doc": { "_id": "id", "_index": "index", "_type": "_doc", "_source": { "field2": "_value2", "foo": "bar" }, "_ingest": { "timestamp": "2017-05-04T22:46:09.674Z" } } }, { "doc": { "_id": "id", "_index": "index", "_type": "_doc", "_source": { "field3": "_value3", "field2": "_value2", "foo": "bar" }, "_ingest": { "timestamp": "2017-05-04T22:46:09.675Z" } } } ] }, { "processor_results": [ { "doc": { "_id": "id", "_index": "index", "_type": "_doc", "_source": { "field2": "_value2", "foo": "rab" }, "_ingest": { "timestamp": "2017-05-04T22:46:09.676Z" } } }, { "doc": { "_id": "id", "_index": "index", "_type": "_doc", "_source": { "field3": "_value3", "field2": "_value2", "foo": "rab" }, "_ingest": { "timestamp": "2017-05-04T22:46:09.677Z" } } } ] } ] } -------------------------------------------------- // TESTRESPONSE[s/"2017-05-04T22:46:09.674Z"/$body.docs.0.processor_results.0.doc._ingest.timestamp/] // TESTRESPONSE[s/"2017-05-04T22:46:09.675Z"/$body.docs.0.processor_results.1.doc._ingest.timestamp/] // TESTRESPONSE[s/"2017-05-04T22:46:09.676Z"/$body.docs.1.processor_results.0.doc._ingest.timestamp/] // TESTRESPONSE[s/"2017-05-04T22:46:09.677Z"/$body.docs.1.processor_results.1.doc._ingest.timestamp/] [[accessing-data-in-pipelines]] == Accessing Data in Pipelines The processors in a pipeline have read and write access to documents that pass through the pipeline. The processors can access fields in the source of a document and the document's metadata fields. [float] [[accessing-source-fields]] === Accessing Fields in the Source Accessing a field in the source is straightforward. You simply refer to fields by their name. For example: [source,js] -------------------------------------------------- { "set": { "field": "my_field", "value": 582.1 } } -------------------------------------------------- // NOTCONSOLE On top of this, fields from the source are always accessible via the `_source` prefix: [source,js] -------------------------------------------------- { "set": { "field": "_source.my_field", "value": 582.1 } } -------------------------------------------------- // NOTCONSOLE [float] [[accessing-metadata-fields]] === Accessing Metadata Fields You can access metadata fields in the same way that you access fields in the source. This is possible because Elasticsearch doesn't allow fields in the source that have the same name as metadata fields. The following example sets the `_id` metadata field of a document to `1`: [source,js] -------------------------------------------------- { "set": { "field": "_id", "value": "1" } } -------------------------------------------------- // NOTCONSOLE The following metadata fields are accessible by a processor: `_index`, `_type`, `_id`, `_routing`. [float] [[accessing-ingest-metadata]] === Accessing Ingest Metadata Fields Beyond metadata fields and source fields, ingest also adds ingest metadata to the documents that it processes. These metadata properties are accessible under the `_ingest` key. Currently ingest adds the ingest timestamp under the `_ingest.timestamp` key of the ingest metadata. The ingest timestamp is the time when Elasticsearch received the index or bulk request to pre-process the document. Any processor can add ingest-related metadata during document processing. Ingest metadata is transient and is lost after a document has been processed by the pipeline. Therefore, ingest metadata won't be indexed. The following example adds a field with the name `received`. The value is the ingest timestamp: [source,js] -------------------------------------------------- { "set": { "field": "received", "value": "{{_ingest.timestamp}}" } } -------------------------------------------------- // NOTCONSOLE Unlike Elasticsearch metadata fields, the ingest metadata field name `_ingest` can be used as a valid field name in the source of a document. Use `_source._ingest` to refer to the field in the source document. Otherwise, `_ingest` will be interpreted as an ingest metadata field. [float] [[accessing-template-fields]] === Accessing Fields and Metafields in Templates A number of processor settings also support templating. Settings that support templating can have zero or more template snippets. A template snippet begins with `{{` and ends with `}}`. Accessing fields and metafields in templates is exactly the same as via regular processor field settings. The following example adds a field named `field_c`. Its value is a concatenation of the values of `field_a` and `field_b`. [source,js] -------------------------------------------------- { "set": { "field": "field_c", "value": "{{field_a}} {{field_b}}" } } -------------------------------------------------- // NOTCONSOLE The following example uses the value of the `geoip.country_iso_code` field in the source to set the index that the document will be indexed into: [source,js] -------------------------------------------------- { "set": { "field": "_index", "value": "{{geoip.country_iso_code}}" } } -------------------------------------------------- // NOTCONSOLE Dynamic field names are also supported. This example sets the field named after the value of `service` to the value of the field `code`: [source,js] -------------------------------------------------- { "set": { "field": "{{service}}", "value": "{{code}}" } } -------------------------------------------------- // NOTCONSOLE [[handling-failure-in-pipelines]] == Handling Failures in Pipelines In its simplest use case, a pipeline defines a list of processors that are executed sequentially, and processing halts at the first exception. This behavior may not be desirable when failures are expected. For example, you may have logs that don't match the specified grok expression. Instead of halting execution, you may want to index such documents into a separate index. To enable this behavior, you can use the `on_failure` parameter. The `on_failure` parameter defines a list of processors to be executed immediately following the failed processor. You can specify this parameter at the pipeline level, as well as at the processor level. If a processor specifies an `on_failure` configuration, whether it is empty or not, any exceptions that are thrown by the processor are caught, and the pipeline continues executing the remaining processors. Because you can define further processors within the scope of an `on_failure` statement, you can nest failure handling. The following example defines a pipeline that renames the `foo` field in the processed document to `bar`. If the document does not contain the `foo` field, the processor attaches an error message to the document for later analysis within Elasticsearch. [source,js] -------------------------------------------------- { "description" : "my first pipeline with handled exceptions", "processors" : [ { "rename" : { "field" : "foo", "target_field" : "bar", "on_failure" : [ { "set" : { "field" : "error", "value" : "field \"foo\" does not exist, cannot rename to \"bar\"" } } ] } } ] } -------------------------------------------------- // NOTCONSOLE The following example defines an `on_failure` block on a whole pipeline to change the index to which failed documents get sent. [source,js] -------------------------------------------------- { "description" : "my first pipeline with handled exceptions", "processors" : [ ... ], "on_failure" : [ { "set" : { "field" : "_index", "value" : "failed-{{ _index }}" } } ] } -------------------------------------------------- // NOTCONSOLE Alternatively instead of defining behaviour in case of processor failure, it is also possible to ignore a failure and continue with the next processor by specifying the `ignore_failure` setting. In case in the example below the field `foo` doesn't exist the failure will be caught and the pipeline continues to execute, which in this case means that the pipeline does nothing. [source,js] -------------------------------------------------- { "description" : "my first pipeline with handled exceptions", "processors" : [ { "rename" : { "field" : "foo", "target_field" : "bar", "ignore_failure" : true } } ] } -------------------------------------------------- // NOTCONSOLE The `ignore_failure` can be set on any processor and defaults to `false`. [float] [[accessing-error-metadata]] === Accessing Error Metadata From Processors Handling Exceptions You may want to retrieve the actual error message that was thrown by a failed processor. To do so you can access metadata fields called `on_failure_message`, `on_failure_processor_type`, and `on_failure_processor_tag`. These fields are only accessible from within the context of an `on_failure` block. Here is an updated version of the example that you saw earlier. But instead of setting the error message manually, the example leverages the `on_failure_message` metadata field to provide the error message. [source,js] -------------------------------------------------- { "description" : "my first pipeline with handled exceptions", "processors" : [ { "rename" : { "field" : "foo", "to" : "bar", "on_failure" : [ { "set" : { "field" : "error", "value" : "{{ _ingest.on_failure_message }}" } } ] } } ] } -------------------------------------------------- // NOTCONSOLE [[ingest-processors]] == Processors All processors are defined in the following way within a pipeline definition: [source,js] -------------------------------------------------- { "PROCESSOR_NAME" : { ... processor configuration options ... } } -------------------------------------------------- // NOTCONSOLE Each processor defines its own configuration parameters, but all processors have the ability to declare `tag` and `on_failure` fields. These fields are optional. A `tag` is simply a string identifier of the specific instantiation of a certain processor in a pipeline. The `tag` field does not affect the processor's behavior, but is very useful for bookkeeping and tracing errors to specific processors. See <> to learn more about the `on_failure` field and error handling in pipelines. The <> can be used to figure out what processors are available in a cluster. The <> will provide a per node list of what processors are available. Custom processors must be installed on all nodes. The put pipeline API will fail if a processor specified in a pipeline doesn't exist on all nodes. If you rely on custom processor plugins make sure to mark these plugins as mandatory by adding `plugin.mandatory` setting to the `config/elasticsearch.yml` file, for example: [source,yaml] -------------------------------------------------- plugin.mandatory: ingest-attachment,ingest-geoip -------------------------------------------------- A node will not start if either of these plugins are not available. The <> can be used to fetch ingest usage statistics, globally and on a per pipeline basis. Useful to find out which pipelines are used the most or spent the most time on preprocessing. [[append-processor]] === Append Processor Appends one or more values to an existing array if the field already exists and it is an array. Converts a scalar to an array and appends one or more values to it if the field exists and it is a scalar. Creates an array containing the provided values if the field doesn't exist. Accepts a single value or an array of values. [[append-options]] .Append Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to be appended to | `value` | yes | - | The value to be appended |====== [source,js] -------------------------------------------------- { "append": { "field": "field1", "value": ["item2", "item3", "item4"] } } -------------------------------------------------- // NOTCONSOLE [[convert-processor]] === Convert Processor Converts an existing field's value to a different type, such as converting a string to an integer. If the field value is an array, all members will be converted. The supported types include: `integer`, `long`, `float`, `double`, `string`, `boolean`, and `auto`. Specifying `boolean` will set the field to true if its string value is equal to `true` (ignore case), to false if its string value is equal to `false` (ignore case), or it will throw an exception otherwise. Specifying `auto` will attempt to convert the string-valued `field` into the closest non-string type. For example, a field whose value is `"true"` will be converted to its respective boolean type: `true`. Do note that float takes precedence of double in `auto`. A value of `"242.15"` will "automatically" be converted to `242.15` of type `float`. If a provided field cannot be appropriately converted, the Convert Processor will still process successfully and leave the field value as-is. In such a case, `target_field` will still be updated with the unconverted field value. [[convert-options]] .Convert Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field whose value is to be converted | `target_field` | no | `field` | The field to assign the converted value to, by default `field` is updated in-place | `type` | yes | - | The type to convert the existing value to | `ignore_missing` | no | `false` | If `true` and `field` does not exist or is `null`, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "convert": { "field" : "foo", "type": "integer" } } -------------------------------------------------- // NOTCONSOLE [[date-processor]] === Date Processor Parses dates from fields, and then uses the date or timestamp as the timestamp for the document. By default, the date processor adds the parsed date as a new field called `@timestamp`. You can specify a different field by setting the `target_field` configuration parameter. Multiple date formats are supported as part of the same date processor definition. They will be used sequentially to attempt parsing the date field, in the same order they were defined as part of the processor definition. [[date-options]] .Date options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to get the date from. | `target_field` | no | @timestamp | The field that will hold the parsed date. | `formats` | yes | - | An array of the expected date formats. Can be a Joda pattern or one of the following formats: ISO8601, UNIX, UNIX_MS, or TAI64N. | `timezone` | no | UTC | The timezone to use when parsing the date. | `locale` | no | ENGLISH | The locale to use when parsing the date, relevant when parsing month names or week days. |====== Here is an example that adds the parsed date to the `timestamp` field based on the `initial_date` field: [source,js] -------------------------------------------------- { "description" : "...", "processors" : [ { "date" : { "field" : "initial_date", "target_field" : "timestamp", "formats" : ["dd/MM/yyyy hh:mm:ss"], "timezone" : "Europe/Amsterdam" } } ] } -------------------------------------------------- // NOTCONSOLE The `timezone` and `locale` processor parameters are templated. This means that their values can be extracted from fields within documents. The example below shows how to extract the locale/timezone details from existing fields, `my_timezone` and `my_locale`, in the ingested document that contain the timezone and locale values. [source,js] -------------------------------------------------- { "description" : "...", "processors" : [ { "date" : { "field" : "initial_date", "target_field" : "timestamp", "formats" : ["ISO8601"], "timezone" : "{{ my_timezone }}", "locale" : "{{ my_locale }}" } } ] } -------------------------------------------------- // NOTCONSOLE [[date-index-name-processor]] === Date Index Name Processor The purpose of this processor is to point documents to the right time based index based on a date or timestamp field in a document by using the <>. The processor sets the `_index` meta field with a date math index name expression based on the provided index name prefix, a date or timestamp field in the documents being processed and the provided date rounding. First, this processor fetches the date or timestamp from a field in the document being processed. Optionally, date formatting can be configured on how the field's value should be parsed into a date. Then this date, the provided index name prefix and the provided date rounding get formatted into a date math index name expression. Also here optionally date formatting can be specified on how the date should be formatted into a date math index name expression. An example pipeline that points documents to a monthly index that starts with a `myindex-` prefix based on a date in the `date1` field: [source,js] -------------------------------------------------- PUT _ingest/pipeline/monthlyindex { "description": "monthly date-time index naming", "processors" : [ { "date_index_name" : { "field" : "date1", "index_name_prefix" : "myindex-", "date_rounding" : "M" } } ] } -------------------------------------------------- // CONSOLE Using that pipeline for an index request: [source,js] -------------------------------------------------- PUT /myindex/_doc/1?pipeline=monthlyindex { "date1" : "2016-04-25T12:02:01.789Z" } -------------------------------------------------- // CONSOLE // TEST[continued] [source,js] -------------------------------------------------- { "_index" : "myindex-2016-04-01", "_type" : "_doc", "_id" : "1", "_version" : 1, "result" : "created", "_shards" : { "total" : 2, "successful" : 1, "failed" : 0 }, "_seq_no" : 0, "_primary_term" : 1 } -------------------------------------------------- // TESTRESPONSE The above request will not index this document into the `myindex` index, but into the `myindex-2016-04-01` index because it was rounded by month. This is because the date-index-name-processor overrides the `_index` property of the document. To see the date-math value of the index supplied in the actual index request which resulted in the above document being indexed into `myindex-2016-04-01` we can inspect the effects of the processor using a simulate request. [source,js] -------------------------------------------------- POST _ingest/pipeline/_simulate { "pipeline" : { "description": "monthly date-time index naming", "processors" : [ { "date_index_name" : { "field" : "date1", "index_name_prefix" : "myindex-", "date_rounding" : "M" } } ] }, "docs": [ { "_source": { "date1": "2016-04-25T12:02:01.789Z" } } ] } -------------------------------------------------- // CONSOLE and the result: [source,js] -------------------------------------------------- { "docs" : [ { "doc" : { "_id" : "_id", "_index" : "", "_type" : "_type", "_source" : { "date1" : "2016-04-25T12:02:01.789Z" }, "_ingest" : { "timestamp" : "2016-11-08T19:43:03.850+0000" } } } ] } -------------------------------------------------- // TESTRESPONSE[s/2016-11-08T19:43:03.850\+0000/$body.docs.0.doc._ingest.timestamp/] The above example shows that `_index` was set to ``. Elasticsearch understands this to mean `2016-04-01` as is explained in the <> [[date-index-name-options]] .Date index name options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to get the date or timestamp from. | `index_name_prefix` | no | - | A prefix of the index name to be prepended before the printed date. | `date_rounding` | yes | - | How to round the date when formatting the date into the index name. Valid values are: `y` (year), `M` (month), `w` (week), `d` (day), `h` (hour), `m` (minute) and `s` (second). | `date_formats` | no | yyyy-MM-dd'T'HH:mm:ss.SSSZ | An array of the expected date formats for parsing dates / timestamps in the document being preprocessed. Can be a Joda pattern or one of the following formats: ISO8601, UNIX, UNIX_MS, or TAI64N. | `timezone` | no | UTC | The timezone to use when parsing the date and when date math index supports resolves expressions into concrete index names. | `locale` | no | ENGLISH | The locale to use when parsing the date from the document being preprocessed, relevant when parsing month names or week days. | `index_name_format` | no | yyyy-MM-dd | The format to be used when printing the parsed date into the index name. An valid Joda pattern is expected here. |====== [[fail-processor]] === Fail Processor Raises an exception. This is useful for when you expect a pipeline to fail and want to relay a specific message to the requester. [[fail-options]] .Fail Options [options="header"] |====== | Name | Required | Default | Description | `message` | yes | - | The error message of the `FailException` thrown by the processor |====== [source,js] -------------------------------------------------- { "fail": { "message": "an error message" } } -------------------------------------------------- // NOTCONSOLE [[foreach-processor]] === Foreach Processor Processes elements in an array of unknown length. All processors can operate on elements inside an array, but if all elements of an array need to be processed in the same way, defining a processor for each element becomes cumbersome and tricky because it is likely that the number of elements in an array is unknown. For this reason the `foreach` processor exists. By specifying the field holding array elements and a processor that defines what should happen to each element, array fields can easily be preprocessed. A processor inside the foreach processor works in the array element context and puts that in the ingest metadata under the `_ingest._value` key. If the array element is a json object it holds all immediate fields of that json object. and if the nested object is a value is `_ingest._value` just holds that value. Note that if a processor prior to the `foreach` processor used `_ingest._value` key then the specified value will not be available to the processor inside the `foreach` processor. The `foreach` processor does restore the original value, so that value is available to processors after the `foreach` processor. Note that any other field from the document are accessible and modifiable like with all other processors. This processor just puts the current array element being read into `_ingest._value` ingest metadata attribute, so that it may be pre-processed. If the `foreach` processor fails to process an element inside the array, and no `on_failure` processor has been specified, then it aborts the execution and leaves the array unmodified. [[foreach-options]] .Foreach Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The array field | `processor` | yes | - | The processor to execute against each field |====== Assume the following document: [source,js] -------------------------------------------------- { "values" : ["foo", "bar", "baz"] } -------------------------------------------------- // NOTCONSOLE When this `foreach` processor operates on this sample document: [source,js] -------------------------------------------------- { "foreach" : { "field" : "values", "processor" : { "uppercase" : { "field" : "_ingest._value" } } } } -------------------------------------------------- // NOTCONSOLE Then the document will look like this after preprocessing: [source,js] -------------------------------------------------- { "values" : ["FOO", "BAR", "BAZ"] } -------------------------------------------------- // NOTCONSOLE Let's take a look at another example: [source,js] -------------------------------------------------- { "persons" : [ { "id" : "1", "name" : "John Doe" }, { "id" : "2", "name" : "Jane Doe" } ] } -------------------------------------------------- // NOTCONSOLE In this case, the `id` field needs to be removed, so the following `foreach` processor is used: [source,js] -------------------------------------------------- { "foreach" : { "field" : "persons", "processor" : { "remove" : { "field" : "_ingest._value.id" } } } } -------------------------------------------------- // NOTCONSOLE After preprocessing the result is: [source,js] -------------------------------------------------- { "persons" : [ { "name" : "John Doe" }, { "name" : "Jane Doe" } ] } -------------------------------------------------- // NOTCONSOLE The wrapped processor can have a `on_failure` definition. For example, the `id` field may not exist on all person objects. Instead of failing the index request, you can use an `on_failure` block to send the document to the 'failure_index' index for later inspection: [source,js] -------------------------------------------------- { "foreach" : { "field" : "persons", "processor" : { "remove" : { "field" : "_value.id", "on_failure" : [ { "set" : { "field", "_index", "value", "failure_index" } } ] } } } } -------------------------------------------------- // NOTCONSOLE In this example, if the `remove` processor does fail, then the array elements that have been processed thus far will be updated. Another advanced example can be found in the {plugins}/ingest-attachment-with-arrays.html[attachment processor documentation]. [[grok-processor]] === Grok Processor Extracts structured fields out of a single text field within a document. You choose which field to extract matched fields from, as well as the grok pattern you expect will match. A grok pattern is like a regular expression that supports aliased expressions that can be reused. This tool is perfect for syslog logs, apache and other webserver logs, mysql logs, and in general, any log format that is generally written for humans and not computer consumption. This processor comes packaged with many https://github.com/elastic/elasticsearch/blob/{branch}/libs/grok/src/main/resources/patterns[reusable patterns]. If you need help building patterns to match your logs, you will find the {kibana-ref}/xpack-grokdebugger.html[Grok Debugger] tool quite useful! The Grok Debugger is an {xpack} feature under the Basic License and is therefore *free to use*. The Grok Constructor at is also a useful tool. [[grok-basics]] ==== Grok Basics Grok sits on top of regular expressions, so any regular expressions are valid in grok as well. The regular expression library is Oniguruma, and you can see the full supported regexp syntax https://github.com/kkos/oniguruma/blob/master/doc/RE[on the Onigiruma site]. Grok works by leveraging this regular expression language to allow naming existing patterns and combining them into more complex patterns that match your fields. The syntax for reusing a grok pattern comes in three forms: `%{SYNTAX:SEMANTIC}`, `%{SYNTAX}`, `%{SYNTAX:SEMANTIC:TYPE}`. The `SYNTAX` is the name of the pattern that will match your text. For example, `3.44` will be matched by the `NUMBER` pattern and `55.3.244.1` will be matched by the `IP` pattern. The syntax is how you match. `NUMBER` and `IP` are both patterns that are provided within the default patterns set. The `SEMANTIC` is the identifier you give to the piece of text being matched. For example, `3.44` could be the duration of an event, so you could call it simply `duration`. Further, a string `55.3.244.1` might identify the `client` making a request. The `TYPE` is the type you wish to cast your named field. `int`, `long`, `double`, `float` and `boolean` are supported types for coercion. For example, you might want to match the following text: [source,txt] -------------------------------------------------- 3.44 55.3.244.1 -------------------------------------------------- You may know that the message in the example is a number followed by an IP address. You can match this text by using the following Grok expression. [source,txt] -------------------------------------------------- %{NUMBER:duration} %{IP:client} -------------------------------------------------- [[using-grok]] ==== Using the Grok Processor in a Pipeline [[grok-options]] .Grok Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to use for grok expression parsing | `patterns` | yes | - | An ordered list of grok expression to match and extract named captures with. Returns on the first expression in the list that matches. | `pattern_definitions` | no | - | A map of pattern-name and pattern tuples defining custom patterns to be used by the current processor. Patterns matching existing names will override the pre-existing definition. | `trace_match` | no | false | when true, `_ingest._grok_match_index` will be inserted into your matched document's metadata with the index into the pattern found in `patterns` that matched. | `ignore_missing` | no | false | If `true` and `field` does not exist or is `null`, the processor quietly exits without modifying the document |====== Here is an example of using the provided patterns to extract out and name structured fields from a string field in a document. [source,js] -------------------------------------------------- { "message": "55.3.244.1 GET /index.html 15824 0.043" } -------------------------------------------------- // NOTCONSOLE The pattern for this could be: [source,txt] -------------------------------------------------- %{IP:client} %{WORD:method} %{URIPATHPARAM:request} %{NUMBER:bytes} %{NUMBER:duration} -------------------------------------------------- Here is an example pipeline for processing the above document by using Grok: [source,js] -------------------------------------------------- { "description" : "...", "processors": [ { "grok": { "field": "message", "patterns": ["%{IP:client} %{WORD:method} %{URIPATHPARAM:request} %{NUMBER:bytes} %{NUMBER:duration}"] } } ] } -------------------------------------------------- // NOTCONSOLE This pipeline will insert these named captures as new fields within the document, like so: [source,js] -------------------------------------------------- { "message": "55.3.244.1 GET /index.html 15824 0.043", "client": "55.3.244.1", "method": "GET", "request": "/index.html", "bytes": 15824, "duration": "0.043" } -------------------------------------------------- // NOTCONSOLE [[custom-patterns]] ==== Custom Patterns and Pattern Files The Grok processor comes pre-packaged with a base set of pattern. These patterns may not always have what you are looking for. Pattern have a very basic format. Each entry describes has a name and the pattern itself. You can add your own patterns to a processor definition under the `pattern_definitions` option. Here is an example of a pipeline specifying custom pattern definitions: [source,js] -------------------------------------------------- { "description" : "...", "processors": [ { "grok": { "field": "message", "patterns": ["my %{FAVORITE_DOG:dog} is colored %{RGB:color}"], "pattern_definitions" : { "FAVORITE_DOG" : "beagle", "RGB" : "RED|GREEN|BLUE" } } } ] } -------------------------------------------------- // NOTCONSOLE [[trace-match]] ==== Providing Multiple Match Patterns Sometimes one pattern is not enough to capture the potential structure of a field. Let's assume we want to match all messages that contain your favorite pet breeds of either cats or dogs. One way to accomplish this is to provide two distinct patterns that can be matched, instead of one really complicated expression capturing the same `or` behavior. Here is an example of such a configuration executed against the simulate API: [source,js] -------------------------------------------------- POST _ingest/pipeline/_simulate { "pipeline": { "description" : "parse multiple patterns", "processors": [ { "grok": { "field": "message", "patterns": ["%{FAVORITE_DOG:pet}", "%{FAVORITE_CAT:pet}"], "pattern_definitions" : { "FAVORITE_DOG" : "beagle", "FAVORITE_CAT" : "burmese" } } } ] }, "docs":[ { "_source": { "message": "I love burmese cats!" } } ] } -------------------------------------------------- // CONSOLE response: [source,js] -------------------------------------------------- { "docs": [ { "doc": { "_type": "_type", "_index": "_index", "_id": "_id", "_source": { "message": "I love burmese cats!", "pet": "burmese" }, "_ingest": { "timestamp": "2016-11-08T19:43:03.850+0000" } } } ] } -------------------------------------------------- // TESTRESPONSE[s/2016-11-08T19:43:03.850\+0000/$body.docs.0.doc._ingest.timestamp/] Both patterns will set the field `pet` with the appropriate match, but what if we want to trace which of our patterns matched and populated our fields? We can do this with the `trace_match` parameter. Here is the output of that same pipeline, but with `"trace_match": true` configured: //// Hidden setup for example: [source,js] -------------------------------------------------- POST _ingest/pipeline/_simulate { "pipeline": { "description" : "parse multiple patterns", "processors": [ { "grok": { "field": "message", "patterns": ["%{FAVORITE_DOG:pet}", "%{FAVORITE_CAT:pet}"], "trace_match": true, "pattern_definitions" : { "FAVORITE_DOG" : "beagle", "FAVORITE_CAT" : "burmese" } } } ] }, "docs":[ { "_source": { "message": "I love burmese cats!" } } ] } -------------------------------------------------- // CONSOLE //// [source,js] -------------------------------------------------- { "docs": [ { "doc": { "_type": "_type", "_index": "_index", "_id": "_id", "_source": { "message": "I love burmese cats!", "pet": "burmese" }, "_ingest": { "_grok_match_index": "1", "timestamp": "2016-11-08T19:43:03.850+0000" } } } ] } -------------------------------------------------- // TESTRESPONSE[s/2016-11-08T19:43:03.850\+0000/$body.docs.0.doc._ingest.timestamp/] In the above response, you can see that the index of the pattern that matched was `"1"`. This is to say that it was the second (index starts at zero) pattern in `patterns` to match. This trace metadata enables debugging which of the patterns matched. This information is stored in the ingest metadata and will not be indexed. [[grok-processor-rest-get]] ==== Retrieving patterns from REST endpoint The Grok Processor comes packaged with its own REST endpoint for retrieving which patterns the processor is packaged with. [source,js] -------------------------------------------------- GET _ingest/processor/grok -------------------------------------------------- // CONSOLE The above request will return a response body containing a key-value representation of the built-in patterns dictionary. [source,js] -------------------------------------------------- { "patterns" : { "BACULA_CAPACITY" : "%{INT}{1,3}(,%{INT}{3})*", "PATH" : "(?:%{UNIXPATH}|%{WINPATH})", ... } -------------------------------------------------- // NOTCONSOLE This can be useful to reference as the built-in patterns change across versions. [[gsub-processor]] === Gsub Processor Converts a string field by applying a regular expression and a replacement. If the field is not a string, the processor will throw an exception. [[gsub-options]] .Gsub Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to apply the replacement to | `pattern` | yes | - | The pattern to be replaced | `replacement` | yes | - | The string to replace the matching patterns with | `target_field` | no | `field` | The field to assign the converted value to, by default `field` is updated in-place | `ignore_missing` | no | `false` | If `true` and `field` does not exist or is `null`, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "gsub": { "field": "field1", "pattern": "\.", "replacement": "-" } } -------------------------------------------------- // NOTCONSOLE [[join-processor]] === Join Processor Joins each element of an array into a single string using a separator character between each element. Throws an error when the field is not an array. [[join-options]] .Join Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to be separated | `separator` | yes | - | The separator character | `target_field` | no | `field` | The field to assign the joined value to, by default `field` is updated in-place |====== [source,js] -------------------------------------------------- { "join": { "field": "joined_array_field", "separator": "-" } } -------------------------------------------------- // NOTCONSOLE [[json-processor]] === JSON Processor Converts a JSON string into a structured JSON object. [[json-options]] .Json Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to be parsed | `target_field` | no | `field` | The field to insert the converted structured object into | `add_to_root` | no | false | Flag that forces the serialized json to be injected into the top level of the document. `target_field` must not be set when this option is chosen. |====== All JSON-supported types will be parsed (null, boolean, number, array, object, string). Suppose you provide this configuration of the `json` processor: [source,js] -------------------------------------------------- { "json" : { "field" : "string_source", "target_field" : "json_target" } } -------------------------------------------------- // NOTCONSOLE If the following document is processed: [source,js] -------------------------------------------------- { "string_source": "{\"foo\": 2000}" } -------------------------------------------------- // NOTCONSOLE after the `json` processor operates on it, it will look like: [source,js] -------------------------------------------------- { "string_source": "{\"foo\": 2000}", "json_target": { "foo": 2000 } } -------------------------------------------------- // NOTCONSOLE If the following configuration is provided, omitting the optional `target_field` setting: [source,js] -------------------------------------------------- { "json" : { "field" : "source_and_target" } } -------------------------------------------------- // NOTCONSOLE then after the `json` processor operates on this document: [source,js] -------------------------------------------------- { "source_and_target": "{\"foo\": 2000}" } -------------------------------------------------- // NOTCONSOLE it will look like: [source,js] -------------------------------------------------- { "source_and_target": { "foo": 2000 } } -------------------------------------------------- // NOTCONSOLE This illustrates that, unless it is explicitly named in the processor configuration, the `target_field` is the same field provided in the required `field` configuration. [[kv-processor]] === KV Processor This processor helps automatically parse messages (or specific event fields) which are of the foo=bar variety. For example, if you have a log message which contains `ip=1.2.3.4 error=REFUSED`, you can parse those automatically by configuring: [source,js] -------------------------------------------------- { "kv": { "field": "message", "field_split": " ", "value_split": "=" } } -------------------------------------------------- // NOTCONSOLE [[kv-options]] .Kv Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to be parsed | `field_split` | yes | - | Regex pattern to use for splitting key-value pairs | `value_split` | yes | - | Regex pattern to use for splitting the key from the value within a key-value pair | `target_field` | no | `null` | The field to insert the extracted keys into. Defaults to the root of the document | `include_keys` | no | `null` | List of keys to filter and insert into document. Defaults to including all keys | `exclude_keys` | no | `null` | List of keys to exclude from document | `ignore_missing` | no | `false` | If `true` and `field` does not exist or is `null`, the processor quietly exits without modifying the document |====== [[lowercase-processor]] === Lowercase Processor Converts a string to its lowercase equivalent. [[lowercase-options]] .Lowercase Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to make lowercase | `target_field` | no | `field` | The field to assign the converted value to, by default `field` is updated in-place | `ignore_missing` | no | `false` | If `true` and `field` does not exist or is `null`, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "lowercase": { "field": "foo" } } -------------------------------------------------- // NOTCONSOLE [[remove-processor]] === Remove Processor Removes existing fields. If one field doesn't exist, an exception will be thrown. [[remove-options]] .Remove Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | Fields to be removed |====== Here is an example to remove a single field: [source,js] -------------------------------------------------- { "remove": { "field": "foo" } } -------------------------------------------------- // NOTCONSOLE To remove multiple fields, you can use the following query: [source,js] -------------------------------------------------- { "remove": { "field": ["foo", "bar"] } } -------------------------------------------------- // NOTCONSOLE [[rename-processor]] === Rename Processor Renames an existing field. If the field doesn't exist or the new name is already used, an exception will be thrown. [[rename-options]] .Rename Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to be renamed | `target_field` | yes | - | The new name of the field | `ignore_missing` | no | `false` | If `true` and `field` does not exist, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "rename": { "field": "foo", "target_field": "foobar" } } -------------------------------------------------- // NOTCONSOLE [[script-processor]] === Script Processor Allows inline and stored scripts to be executed within ingest pipelines. See <> to learn more about writing scripts. The Script Processor leverages caching of compiled scripts for improved performance. Since the script specified within the processor is potentially re-compiled per document, it is important to understand how script caching works. To learn more about caching see <>. [[script-options]] .Script Options [options="header"] |====== | Name | Required | Default | Description | `lang` | no | "painless" | The scripting language | `id` | no | - | The stored script id to refer to | `source` | no | - | An inline script to be executed | `params` | no | - | Script Parameters |====== One of `id` or `source` options must be provided in order to properly reference a script to execute. You can access the current ingest document from within the script context by using the `ctx` variable. The following example sets a new field called `field_a_plus_b_times_c` to be the sum of two existing numeric fields `field_a` and `field_b` multiplied by the parameter param_c: [source,js] -------------------------------------------------- { "script": { "lang": "painless", "source": "ctx.field_a_plus_b_times_c = (ctx.field_a + ctx.field_b) * params.param_c", "params": { "param_c": 10 } } } -------------------------------------------------- // NOTCONSOLE It is possible to use the Script Processor to manipulate document metadata like `_index` and `_type` during ingestion. Here is an example of an Ingest Pipeline that renames the index and type to `my_index` no matter what was provided in the original index request: [source,js] -------------------------------------------------- PUT _ingest/pipeline/my_index { "description": "use index:my_index and type:_doc", "processors": [ { "script": { "source": """ ctx._index = 'my_index'; ctx._type = '_doc'; """ } } ] } -------------------------------------------------- // CONSOLE Using the above pipeline, we can attempt to index a document into the `any_index` index. [source,js] -------------------------------------------------- PUT any_index/_doc/1?pipeline=my_index { "message": "text" } -------------------------------------------------- // CONSOLE // TEST[continued] The response from the above index request: [source,js] -------------------------------------------------- { "_index": "my_index", "_type": "_doc", "_id": "1", "_version": 1, "result": "created", "_shards": { "total": 2, "successful": 1, "failed": 0 }, "_seq_no": 0, "_primary_term": 1, } -------------------------------------------------- // TESTRESPONSE In the above response, you can see that our document was actually indexed into `my_index` instead of `any_index`. This type of manipulation is often convenient in pipelines that have various branches of transformation, and depending on the progress made, indexed into different indices. [[set-processor]] === Set Processor Sets one field and associates it with the specified value. If the field already exists, its value will be replaced with the provided one. [[set-options]] .Set Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to insert, upsert, or update | `value` | yes | - | The value to be set for the field | `override`| no | true | If processor will update fields with pre-existing non-null-valued field. When set to `false`, such fields will not be touched. |====== [source,js] -------------------------------------------------- { "set": { "field": "field1", "value": 582.1 } } -------------------------------------------------- // NOTCONSOLE [[split-processor]] === Split Processor Splits a field into an array using a separator character. Only works on string fields. [[split-options]] .Split Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to split | `separator` | yes | - | A regex which matches the separator, eg `,` or `\s+` | `target_field` | no | `field` | The field to assign the split value to, by default `field` is updated in-place | `ignore_missing` | no | `false` | If `true` and `field` does not exist, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "split": { "field": "my_field", "separator": "\\s+" <1> } } -------------------------------------------------- // NOTCONSOLE <1> Treat all consecutive whitespace characters as a single separator [[sort-processor]] === Sort Processor Sorts the elements of an array ascending or descending. Homogeneous arrays of numbers will be sorted numerically, while arrays of strings or heterogeneous arrays of strings + numbers will be sorted lexicographically. Throws an error when the field is not an array. [[sort-options]] .Sort Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to be sorted | `order` | no | `"asc"` | The sort order to use. Accepts `"asc"` or `"desc"`. | `target_field` | no | `field` | The field to assign the sorted value to, by default `field` is updated in-place |====== [source,js] -------------------------------------------------- { "sort": { "field": "field_to_sort", "order": "desc" } } -------------------------------------------------- // NOTCONSOLE [[trim-processor]] === Trim Processor Trims whitespace from field. NOTE: This only works on leading and trailing whitespace. [[trim-options]] .Trim Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The string-valued field to trim whitespace from | `target_field` | no | `field` | The field to assign the trimmed value to, by default `field` is updated in-place | `ignore_missing` | no | `false` | If `true` and `field` does not exist, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "trim": { "field": "foo" } } -------------------------------------------------- // NOTCONSOLE [[uppercase-processor]] === Uppercase Processor Converts a string to its uppercase equivalent. [[uppercase-options]] .Uppercase Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to make uppercase | `target_field` | no | `field` | The field to assign the converted value to, by default `field` is updated in-place | `ignore_missing` | no | `false` | If `true` and `field` does not exist or is `null`, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "uppercase": { "field": "foo" } } -------------------------------------------------- // NOTCONSOLE [[dot-expand-processor]] === Dot Expander Processor Expands a field with dots into an object field. This processor allows fields with dots in the name to be accessible by other processors in the pipeline. Otherwise these <> can't be accessed by any processor. [[dot-expender-options]] .Dot Expand Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to expand into an object field | `path` | no | - | The field that contains the field to expand. Only required if the field to expand is part another object field, because the `field` option can only understand leaf fields. |====== [source,js] -------------------------------------------------- { "dot_expander": { "field": "foo.bar" } } -------------------------------------------------- // NOTCONSOLE For example the dot expand processor would turn this document: [source,js] -------------------------------------------------- { "foo.bar" : "value" } -------------------------------------------------- // NOTCONSOLE into: [source,js] -------------------------------------------------- { "foo" : { "bar" : "value" } } -------------------------------------------------- // NOTCONSOLE If there is already a `bar` field nested under `foo` then this processor merges the `foo.bar` field into it. If the field is a scalar value then it will turn that field into an array field. For example, the following document: [source,js] -------------------------------------------------- { "foo.bar" : "value2", "foo" : { "bar" : "value1" } } -------------------------------------------------- // NOTCONSOLE is transformed by the `dot_expander` processor into: [source,js] -------------------------------------------------- { "foo" : { "bar" : ["value1", "value2"] } } -------------------------------------------------- // NOTCONSOLE If any field outside of the leaf field conflicts with a pre-existing field of the same name, then that field needs to be renamed first. Consider the following document: [source,js] -------------------------------------------------- { "foo": "value1", "foo.bar": "value2" } -------------------------------------------------- // NOTCONSOLE Then the `foo` needs to be renamed first before the `dot_expander` processor is applied. So in order for the `foo.bar` field to properly be expanded into the `bar` field under the `foo` field the following pipeline should be used: [source,js] -------------------------------------------------- { "processors" : [ { "rename" : { "field" : "foo", "target_field" : "foo.bar"" } }, { "dot_expander": { "field": "foo.bar" } } ] } -------------------------------------------------- // NOTCONSOLE The reason for this is that Ingest doesn't know how to automatically cast a scalar field to an object field. [[urldecode-processor]] === URL Decode Processor URL-decodes a string [[urldecode-options]] .URL Decode Options [options="header"] |====== | Name | Required | Default | Description | `field` | yes | - | The field to decode | `target_field` | no | `field` | The field to assign the converted value to, by default `field` is updated in-place | `ignore_missing` | no | `false` | If `true` and `field` does not exist or is `null`, the processor quietly exits without modifying the document |====== [source,js] -------------------------------------------------- { "urldecode": { "field": "my_url_to_decode" } } -------------------------------------------------- // NOTCONSOLE