Throttle filter plugin

edit
  • Plugin version: v4.0.2
  • Released on: 2017-06-23
  • Changelog

Getting Help

edit

For questions about the plugin, open a topic in the Discuss forums. For bugs or feature requests, open an issue in Github. For the list of Elastic supported plugins, please consult the Elastic Support Matrix.

Description

edit

The throttle filter is for throttling the number of events. The filter is configured with a lower bound, the "before_count", and upper bound, the "after_count", and a period of time. All events passing through the filter will be counted based on their key and the event timestamp. As long as the count is less than the "before_count" or greater than the "after_count", the event will be "throttled" which means the filter will be considered successful and any tags or fields will be added (or removed).

The plugin is thread-safe and properly tracks past events.

For example, if you wanted to throttle events so you only receive an event after 2 occurrences and you get no more than 3 in 10 minutes, you would use the configuration:

    period => 600
    max_age => 1200
    before_count => 3
    after_count => 5

Which would result in:

event 1 - throttled (successful filter, period start)
event 2 - throttled (successful filter)
event 3 - not throttled
event 4 - not throttled
event 5 - not throttled
event 6 - throttled (successful filter)
event 7 - throttled (successful filter)
event x - throttled (successful filter)
period end
event 1 - throttled (successful filter, period start)
event 2 - throttled (successful filter)
event 3 - not throttled
event 4 - not throttled
event 5 - not throttled
event 6 - throttled (successful filter)
...

Another example is if you wanted to throttle events so you only receive 1 event per hour, you would use the configuration:

    period => 3600
    max_age => 7200
    before_count => -1
    after_count => 1

Which would result in:

event 1 - not throttled (period start)
event 2 - throttled (successful filter)
event 3 - throttled (successful filter)
event 4 - throttled (successful filter)
event x - throttled (successful filter)
period end
event 1 - not throttled (period start)
event 2 - throttled (successful filter)
event 3 - throttled (successful filter)
event 4 - throttled (successful filter)
...

A common use case would be to use the throttle filter to throttle events before 3 and after 5 while using multiple fields for the key and then use the drop filter to remove throttled events. This configuration might appear as:

    filter {
      throttle {
        before_count => 3
        after_count => 5
        period => 3600
        max_age => 7200
        key => "%{host}%{message}"
        add_tag => "throttled"
      }
      if "throttled" in [tags] {
        drop { }
      }
    }

Another case would be to store all events, but only email non-throttled events so the op’s inbox isn’t flooded with emails in the event of a system error. This configuration might appear as:

    filter {
      throttle {
        before_count => 3
        after_count => 5
        period => 3600
        max_age => 7200
        key => "%{message}"
        add_tag => "throttled"
      }
    }
    output {
      if "throttled" not in [tags] {
        email {
          from => "[email protected]"
          subject => "Production System Alert"
          to => "[email protected]"
          via => "sendmail"
          body => "Alert on %{host} from path %{path}:\n\n%{message}"
          options => { "location" => "/usr/sbin/sendmail" }
        }
      }
      elasticsearch_http {
        host => "localhost"
        port => "19200"
      }
    }

When an event is received, the event key is stored in a key_cache. The key references a timeslot_cache. The event is allocated to a timeslot (created dynamically) based on the timestamp of the event. The timeslot counter is incremented. When the next event is received (same key), within the same "period", it is allocated to the same timeslot. The timeslot counter is incremented once again.

The timeslot expires if the maximum age has been exceeded. The age is calculated based on the latest event timestamp and the max_age configuration option.

        ---[::.. DESIGN ..::]---

+- [key_cache] -+  +-- [timeslot_cache] --+
|               |  | @created: 1439839636 |
                   | @latest:  1439839836 |
   [a.b.c]  =>     +----------------------+
                   | [1439839636] => 1    |
                   | [1439839736] => 3    |
                   | [1439839836] => 2    |
                   +----------------------+

                   +-- [timeslot_cache] --+
                   | @created: eeeeeeeeee |
                   | @latest:  llllllllll |
   [x.y.z]  =>     +----------------------+
                   | [0000000060] => x    |
                   | [0000000120] => y    |
|               |  | [..........] => N    |
+---------------+  +----------------------+

Frank de Jong (@frapex) Mike Pilone (@mikepilone)

only update if greater than current

Throttle Filter Configuration Options

edit

This plugin supports the following configuration options plus the Common Options described later.

Setting Input type Required

after_count

number

No

before_count

number

No

key

string

Yes

max_age

number

No

max_counters

number

No

period

string

No

Also see Common Options for a list of options supported by all filter plugins.

 

after_count

edit
  • Value type is number
  • Default value is -1

Events greater than this count will be throttled. Setting this value to -1, the default, will cause no events to be throttled based on the upper bound.

before_count

edit
  • Value type is number
  • Default value is -1

Events less than this count will be throttled. Setting this value to -1, the default, will cause no events to be throttled based on the lower bound.

key

edit
  • This is a required setting.
  • Value type is string
  • There is no default value for this setting.

The key used to identify events. Events with the same key are grouped together. Field substitutions are allowed, so you can combine multiple fields.

max_age

edit
  • Value type is number
  • Default value is 3600

The maximum age of a timeslot. Higher values allow better tracking of an asynchronous flow of events, but require more memory. As a rule of thumb you should set this value to at least twice the period. Or set this value to period + maximum time offset between unordered events with the same key. Values below the specified period give unexpected results if unordered events are processed simultaneously.

max_counters

edit
  • Value type is number
  • Default value is 100000

The maximum number of counters to store before decreasing the maximum age of a timeslot. Setting this value to -1 will prevent an upper bound with no constraint on the number of counters. This configuration value should only be used as a memory control mechanism and can cause early counter expiration if the value is reached. It is recommended to leave the default value and ensure that your key is selected such that it limits the number of counters required (i.e. don’t use UUID as the key).

period

edit
  • Value type is string
  • Default value is "60"

The period in seconds after the first occurrence of an event until a new timeslot is created. This period is tracked per unique key and per timeslot. Field substitutions are allowed in this value. This allows you to specify that certain kinds of events throttle for a specific period of time.

Common Options

edit

The following configuration options are supported by all filter plugins:

add_field

edit
  • Value type is hash
  • Default value is {}

If this filter is successful, add any arbitrary fields to this event. Field names can be dynamic and include parts of the event using the %{field}.

Example:

filter {
  PLUGIN_NAME {
    add_field => { "foo_%{somefield}" => "Hello world, from %{host}" }
  }
}
# You can also add multiple fields at once:
filter {
  PLUGIN_NAME {
    add_field => {
      "foo_%{somefield}" => "Hello world, from %{host}"
      "new_field" => "new_static_value"
    }
  }
}

If the event has field "somefield" == "hello" this filter, on success, would add field foo_hello if it is present, with the value above and the %{host} piece replaced with that value from the event. The second example would also add a hardcoded field.

add_tag

edit
  • Value type is array
  • Default value is []

If this filter is successful, add arbitrary tags to the event. Tags can be dynamic and include parts of the event using the %{field} syntax.

Example:

filter {
  PLUGIN_NAME {
    add_tag => [ "foo_%{somefield}" ]
  }
}
# You can also add multiple tags at once:
filter {
  PLUGIN_NAME {
    add_tag => [ "foo_%{somefield}", "taggedy_tag"]
  }
}

If the event has field "somefield" == "hello" this filter, on success, would add a tag foo_hello (and the second example would of course add a taggedy_tag tag).

enable_metric

edit
  • Value type is boolean
  • Default value is true

Disable or enable metric logging for this specific plugin instance by default we record all the metrics we can, but you can disable metrics collection for a specific plugin.

  • Value type is string
  • There is no default value for this setting.

Add a unique ID to the plugin instance, this ID is used for tracking information for a specific configuration of the plugin.

output {
 stdout {
   id => "ABC"
 }
}

If you don’t explicitely set this variable Logstash will generate a unique name.

periodic_flush

edit
  • Value type is boolean
  • Default value is false

Call the filter flush method at regular interval. Optional.

remove_field

edit
  • Value type is array
  • Default value is []

If this filter is successful, remove arbitrary fields from this event. Fields names can be dynamic and include parts of the event using the %{field} Example:

filter {
  PLUGIN_NAME {
    remove_field => [ "foo_%{somefield}" ]
  }
}
# You can also remove multiple fields at once:
filter {
  PLUGIN_NAME {
    remove_field => [ "foo_%{somefield}", "my_extraneous_field" ]
  }
}

If the event has field "somefield" == "hello" this filter, on success, would remove the field with name foo_hello if it is present. The second example would remove an additional, non-dynamic field.

remove_tag

edit
  • Value type is array
  • Default value is []

If this filter is successful, remove arbitrary tags from the event. Tags can be dynamic and include parts of the event using the %{field} syntax.

Example:

filter {
  PLUGIN_NAME {
    remove_tag => [ "foo_%{somefield}" ]
  }
}
# You can also remove multiple tags at once:
filter {
  PLUGIN_NAME {
    remove_tag => [ "foo_%{somefield}", "sad_unwanted_tag"]
  }
}

If the event has field "somefield" == "hello" this filter, on success, would remove the tag foo_hello if it is present. The second example would remove a sad, unwanted tag as well.