Common filter plugins
This page contains a list of common filter plugins.
mutate
You can use the mutate
filter to change the data type of a field. For example, you can use the mutate
filter if you’re sending events to Lucenia and you need to change the data type of a field to match any existing mappings.
To convert the quantity
field from a string
type to an integer
type:
input {
http {
host => "127.0.0.1"
port => 8080
}
}
filter {
mutate {
convert => {"quantity" => "integer"}
}
}
output {
file {
path => "output.txt"
}
}
Sample output
You can see that the type of the quantity
field is changed from a string
to an integer
.
{
"quantity" => 3,
"host" => "127.0.0.1",
"@timestamp" => 2021-05-23T19:02:08.026Z,
"amount" => 10,
"@version" => "1",
"headers" => {
"request_path" => "/",
"connection" => "keep-alive",
"content_length" => "41",
"http_user_agent" => "PostmanRuntime/7.26.8",
"request_method" => "PUT",
"cache_control" => "no-cache",
"http_accept" => "*/*",
"content_type" => "application/json",
"http_version" => "HTTP/1.1",
"http_host" => "127.0.0.1:8080",
"accept_encoding" => "gzip, deflate, br",
"postman_token" => "ffd1cdcb-7a1d-4d63-90f8-0f2773069205"
}
}
Other data types you can convert to are float
, string
, and boolean
values. If you pass in an array, the mutate
filter converts all the elements in the array. If you pass a string
like “world” to cast to an integer
type, the result is 0 and Logstash continues processing events.
Logstash supports a few common options for all filter plugins:
Option | Description |
---|---|
add_field | Adds one or more fields to the event. |
remove_field | Removes one or more events from the field. |
add_tag | Adds one or more tags to the event. You can use tags to perform conditional processing on events depending on which tags they contain. |
remove_tag | Removes one or more tags from the event. |
For example, you can remove the host
field from the event:
input {
http {
host => "127.0.0.1"
port => 8080
}
}
filter {
mutate {
remove_field => {"host"}
}
}
output {
file {
path => "output.txt"
}
}
grok
With the grok
filter, you can parse unstructured data and and structure it into fields. The grok
filter uses text patterns to match text in your logs. You can think of text patterns as variables containing regular expressions.
The format of a text pattern is as follows:
%{SYNTAX:SEMANTIC}
SYNTAX
is the format a piece of text should be in for the pattern to match. You can enter any of grok
’s predefined patterns. For example, you can use the email identifier to match an email address from a given piece of text.
SEMANTIC
is an arbitrary name for the matched text. For example, if you’re using the email identifier syntax, you can name it “email.”
The following request consists of the IP address of the visitor, name of the visitor, the timestamp of the request, the HTTP verb and URL, the HTTP status code, and the number of bytes:
184.252.108.229 - joe [20/Sep/2017:13:22:22 +0200] GET /products/view/123 200 12798
To split this request into different fields:
filter {
grok {
match => { "message" => " %{IP: ip_address} %{USER:identity}
%{USER:auth} \[%{HTTPDATE:reg_ts}\]
\"%{WORD:http_verb}
%{URIPATHPARAM: req_path}
\" %{INT:http_status:int}
%{INT:num_bytes:int}"}
}
}
where:
IP
: matches the IP address field.USER
: matches the user name.WORD
: matches the HTTP verb.URIPATHPARAM
: matches the URI path.INT
: matches the HTTP status field.INT
: matches the number of bytes.
This is what the event looks like after the grok
filter breaks it down into individual fields:
ip_address: 184.252.108.229
identity: joe
reg_ts: 20/Sep/2017:13:22:22 +0200
http_verb:GET
req_path: /products/view/123
http_status: 200
num_bytes: 12798
For common log formats, you use the predefined patterns defined here—Logstash patterns. You can make any adjustments to the results with the mutate
filter.