Versions | v0.12 (td-agent2)

parser Filter Plugin

The filter_parser filter plugin “parses” string field in event records and mutates its event record with parsed result.

Table of Contents

Example Configurations

filter_parser is included in Fluentd’s core since v0.12.29. No installation required. If you want to use filter_parser with lower fluentd versions, need to install fluent-plugin-parser.

filter_parser has just same with in_tail about format and time_format:

<filter foo.bar>
  @type parser
  format /^(?<host>[^ ]*) [^ ]* (?<user>[^ ]*) \[(?<time>[^\]]*)\] "(?<method>\S+)(?: +(?<path>[^ ]*) +\S*)?" (?<code>[^ ]*) (?<size>[^ ]*)$/
  time_format %d/%b/%Y:%H:%M:%S %z
  key_name message
</filter>

filter_parser uses built-in parser plugins and your own customized parser plugin, so you can re-use pre-defined format like apache, json and etc. See document page for more details: Parser Plugin Overview

Parameters

format

This is required parameter. Specify parser format or regexp pattern.

key_name

This is required parameter. Specify field name in the record to parse.

reserve_data

Keep original key-value pair in parsed result.

<filter foo.bar>
  @type parser
  format json
  key_name log
  reserve_data true
</filter>

With above configuration, result is below:

# input data:  {"key":"value","log":"{\"user\":1,\"num\":2}"}
# output data: {"key":"value","log":"{\"user\":1,\"num\":2}","user":1,"num":2}

Without reserve_data, result is below

# input data:  {"key":"value","log":"{\"user\":1,\"num\":2}"}
# output data: {"user":1,"num":2}

suppress_parse_error_log

If true, a plugin suppresses pattern not match warning log. Default is false.

This parameter is useful for parsing mixed logs and you want to ignore non target lines.

ignore_key_not_exist

Ignore “key not exist” log. Default is false.

Useful case is same with suppress_parse_error_log.

replace_invalid_sequence

If true, invalid string is replaced with safe characters and re-parse it. Default is false.

inject_key_prefix

Store parsed values with specified key name prefix. Default is nil.

<filter foo.bar>
  @type parser
  format json
  key_name log
  reserve_data true
  inject_key_prefix data.
</filter>

With above configuration, result is below:

# input data:  {"log": "{\"user\":1,\"num\":2}"}
# output data: {"log":"{\"user\":1,\"num\":2}","data.user":1, "data.num":2}

hash_value_field

Store parsed values as a hash value in a field. Default is nil.

<filter foo.bar>
  @type parser
  format json
  key_name log
  hash_value_field parsed
</filter>

With above configuration, result is below:

# input data:  {"log": "{\"user\":1,\"num\":2}"}
# output data: {"parsed":{"user":1,"num":2}}

time_parse

If false, time parsing is disabled in the parser. Default is true.

emit_invalid_record_to_error

Emit invalid record to @ERROR label. Default is false. Invalid cases are

  • key not exist
  • format is not matched
  • unexpected error

You can rescue unexpected format logs in @ERROR label.

Learn More

Last updated: 2017-03-16 10:36:00 UTC

Versions | v0.12 (td-agent2)

If this article is incorrect or outdated, or omits critical information, please let us know. Fluentd is a open source project under Cloud Native Computing Foundation (CNCF), originally invented by Treasure Data, Inc. All components are available under the Apache 2 License.

Interested in the Fluentd Newsletters?