Fluent output filter plugin for parsing key/value fields in records based on <key>=<value> pattern.
Installation
Use RubyGems:
gem install fluent-plugin-fields-parser
Configuration
<match pattern>
type fields_parser
remove_tag_prefix raw
add_tag_prefix parsed
strict_key_value false
</match>
If following record is passed:
{"message": "Audit log user=Johny action='add-user' result=success" }
then you will get a new record:
{
"message": "Audit log username=Johny action='add-user' result=success",
"user": "Johny",
"action": "add-user",
"result": "success"
}
Parameter parse_key
For configuration
<match pattern>
type fields_parser
parse_key log_message
</match>
it parses key "log_message" instead of default key message
.
Parameter fields_key
Configuration
<match pattern>
type fields_parser
parse_key log_message
fields_key fields
</match>
For input like:
{
"log_message": "Audit log username=Johny action='add-user' result=success",
}
it adds parsed fields into defined key.
{
"log_message": "Audit log username=Johny action='add-user' result=success",
"fields": {"user": "Johny", "action": "add-user", "result": "success"}
}
(It adds new keys into top-level record by default.)
Parameter pattern
You can define custom pattern (regexp) for seaching keys/values.
Configuration
<match pattern>
type fields_parser
pattern (\w+):(\d+)
</match>
For input like:
{ "message": "data black:54 white:55 red:10"}
it returns:
{ "message": "data black:54 white=55 red=10",
"black": "54", "white": "55", "red": "10"
}
Tag prefix
You cat add and/or remove tag prefix using Configuration parameters
<match pattern>
type fields_parser
remove_tag_prefix raw
add_tag_prefix parsed
</match>
If it matched tag "raw.some.record", then it emits tag "parsed.some.record".
Parameter strict_key_value
<match pattern>
type fields_parser
strict_key_value true
</match>
If strict_key_value
is set to true
, the parser will use the ruby logfmt
parser which will parse the log
message based on the popular logfmt key/value
format. Do note that this parser will create Fixnum and Float type values
when it parses integer and float values.
All information provided in the log message must be in a strict key=value format. For example, if following record is passed:
{"message": "msg=\"Audit log\" user=Johnny action=\"add-user\" result=success iVal=23 fVal=1.02 bVal=true" }
then you will get a new record:
{
"message": "msg=\"Audit log\" user=Johnny action=\"add-user\" result=success iVal=23 fVal=1.02 bVal=true",
"msg": "Audit log",
"user": "Johnny",
"action": "add-user",
"result": "success",
"iVal": 23,
"fVal": 1.02,
"bVal": "true"
}