Difference between revisions of "Plugin:Write Riemann"

Revision as of 14:27, 18 June 2014

The Write Riemann plugin stores values in Riemann, a stream processing and monitoring system.

Riemann is a stream processing system that can be used for querying, manipulating and visualizing value streams. This makes Riemann a powerful and useful companion for collectd because it allows more sophisticated alerting than is possible with collectd's built-in thresholds.

The Write Riemann plugin uses UDP packets or a TCP connection to send data to Riemann. The data is binary encoded using Protocol Buffers (protobuf), see section #Example data below.

You can configure multiple Node blocks to send data to multiple instances of Riemann. You can use the name specified in the block to send data selectively, using filters. For example, if you configure the connection to an instance of Riemann with

<Node "foobar">
  # ...
</Node>

you can write to this instance only using the block syntax of the Write target:

<Target "write">
  Plugin "write_riemann/foobar"
</Target>

Synopsis

<Plugin "write_riemann">
 <Node "example">
   Host "localhost"
   Port "5555"
   Protocol UDP
   StoreRates true
   AlwaysAppendDS false
 </Node>
 Tag "foobar"
</Plugin>

This will register a writer with the name

write_riemann/localhost

Example data

Data is sent as Msg objects, encoded in a protobuf. You can find the definition of the message types in src/riemann.proto.

events {
  time: 1358875608
  service: "cpu-0/cpu-idle"
  host: "server.example.com"
  attribute {
    key: "plugin"
    value: "cpu"
  }
  attribute {
    key: "plugin_instance"
    value: "0"
  }
  attribute {
    key: "type"
    value: "cpu"
  }
  attribute {
    key: "type_instance"
    value: "idle"
  }
  attribute {
    key: "ds_type"
    value: "derive:rate"
  }
  attribute {
    key: "ds_name"
    value: "value"
  }
  attribute {
    key: "ds_index"
    value: "0"
  }
  ttl: 20
  metric_d: 66.600990955859146
}

Dependencies

• The <google/protobuf-c/protobuf-c.h> header file.

Caveats

Precision

Currently, Riemann provides a precision of one second. We recommend to use at least a two-second interval when sending data to Riemann, to avoid "duplicate" metrics when the time is rounded to the nearest second.

See also

• Riemann website