CircleCI

fluent-plugin-splunk-hec

Fluentd output plugin to send events and metrics to Splunk over the HEC (HTTP Event Collector) API.

Installation

RubyGems

$ gem install fluent-plugin-splunk-hec

Bundler

Add following line to your Gemfile:

gem "fluent-plugin-splunk-hec"

And then execute:

$ bundle

Configuration

Examples

Example 1: Minimum Configs

<match **>
  @type splunk_hec
  hec_host 12.34.56.78
  hec_port 8088
  hec_token 00000000-0000-0000-0000-000000000000
</match>

This example is very basic, it just tells the plugin to send events to Splunk HEC on https://12.34.56.78:8088 (https is the default protocol), using the HEC token 00000000-0000-0000-0000-000000000000. It will use whatever index, source, sourcetype are configured in HEC. And the host of each event is the hostname of the machine which running fluentd.

Example 2: Overwrite HEC defaults

<match **>
  @type splunk_hec
  hec_host 12.34.56.78
  hec_port 8088
  hec_token 00000000-0000-0000-0000-000000000000

  index awesome
  source ${tag}
  sourcetype _json
</match>

This configuration will

  • send all events to the awesome index, and
  • set their source to the event tags. ${tag} is a special value which will be replaced by the event tags, and
  • set their sourcetype to _json.

Sometimes you want to use the values from the input event for these parameters, this is where the *_key parameters help.

<match **>
  ...omitting other parameters...

  source_key file_path
</match>

In the second example (in order to keep it concise, we just omitted the repeating parameters, and we will keep doing so in the following examples), it uses the source_key config to set the source of event to the value of the event's file_path field. Given an input event like

{"file_path": "/var/log/splunk.log", "message": "This is an exmaple.", "level": "info"}

Then the source for this event will be "/var/log/splunk.log". And the "file_path" field will be removed from the input event, so what you will eventually get ingested in Splunk is

{"message": "This is an exmaple.", "level": "info"}

If you want to keep "file_path" in the event, you can use keep_keys.

Besides source_key there are also other *_key parameters, check the parameters details below.

Example 3: Sending metrics

Metrics is avaialble since Splunk 7.0.0, you can use this output plugin to send events as metrics to a Splunk metric index by setting data_type to "metric".

<match **>
  @type splunk_hec
  data_type metric
  hec_host 12.34.56.78
  hec_port 8088
  hec_token 00000000-0000-0000-0000-000000000000
</match>

With this configuration, the plugin will treat each input event as a collection of metrics, i.e. each key-varlue pair in the event is a metric name-value pair. For example, given an input event like

{"cpu/usage": 0.5, "cpu/rate": 10, "memory/usage": 100, "memory/rss": 90}

then 4 metrics will be sent to splunk.

If the input events are not like this, instead they have the metric name and metric value as properties of the event. Then you can use metric_name_key and metric_value_key. Given an input event like

{"metric": "cpu/usage", "value": 0.5, "app": "web_ui"}

You should change the configuration to

<match **>
  @type splunk_hec
  data_type metric
  hec_host 12.34.56.78
  hec_port 8088
  hec_token 00000000-0000-0000-0000-000000000000

  metric_name_key metric
  metric_value_key value
</match>

All other properties of the input (in this example, "app"), will be sent as dimensions of the metric. You can use the <fields> section to customize the dimensions.

Parameters

@type

This value must be splunk_hec.

protocol (enum) (optional)

Protocol to use to call HEC API.

Available values: http, https

Default value: https.

hec_host (string) (required)

The hostname/IP to HEC, or HEC load balancer.

hec_port (integer) (optional)

The port number to HEC, or HEC load balancer.

Default value: 8088.

hec_token (string) (required)

The HEC token.

index (string) (optional)

The Splunk index to index events. When not set, will be decided by HEC. This is exclusive with index_key.

index_key (string) (optional)

Field name to contain Splunk index name. This is exclusive with index.

host (string) (optional)

The host field for events. This is exclusive with host_key.

Default value: the hostname of the host machine.

host_key (string) (optional)

Field name to contain host. This is exclusive with host.

source (string) (optional)

The source field for events, when not set, will be decided by HEC. This is exclusive with source_key.

source_key (string) (optional)

Field name to contain source. This is exclusive with source.

sourcetype (string) (optional)

The sourcetype field for events, when not set, will be decided by HEC. This is exclusive with sourcetype_key.

sourcetype_key (string) (optional)

Field name to contain sourcetype. This is exclusive with sourcetype.

metrics_from_event (bool) (optional)

When data_type is set to "metric", by default it will treat every key-value pair in the input event as a metric name-value pair. Set metrics_from_event to false to disable this behavior and use metric_name_key and metric_value_key to define metrics.

Default value: true.

metric_name_key (string) (optional)

Field name to contain metric name. This is exclusive with metrics_from_event, when this is set, metrics_from_event will be set to false.

metric_value_key (string) (optional)

Field name to contain metric value, this is required when metric_name_key is set.

keep_keys (bool) (optional)

By default, all the fields used by the *_key parameters will be removed from the original input events. To change this behavior, set this parameter to true.

Default value: true.

coerce_to_utf8 (bool) (optional)

Whether to allow non-UTF-8 characters in user logs. If set to true, any non-UTF-8 character would be replaced by the string specified by non_utf8_replacement_string. If set to false, any non-UTF-8 character would trigger the plugin to error out.

Default value: true.

non_utf8_replacement_string (string) (optional)

If coerce_to_utf8 is set to true, any non-UTF-8 character would be replaced by the string specified here.

Default value: ' '.

<fields> section (optional) (single)

Depending on the value of data_type parameter, the parameters inside <fields> section have different meanings. Despite the meaning, the syntax for parameters is unique.

When data_type is event

In this case, parameters inside <fields> will be used as indexed fields. And these fields will be removed from the original input events. Please see the "Add a "fields" property at the top JSON level" here for details. Given we have configuration like

<match **>
  @type splunk_hec
  ...omitting other parameters...

  <fields>
    file
    level
    app applicatioin
  </fields>
</match>

and an input event like

{"application": "webServer", "file": "server.rb", "lineNo": 100, "level": "info", "message": "Request finished in 30ms."}

Then the HEC request JSON payload will be:

{
   // omitting other fields
   // ...
   "event": "{\"lineNo\": 100, \"message\": \"Request finished in 30ms.\"}",
   "fields": {
     "file": "server.rb",
     "level": "info",
     "app": "webServer"
   }
}

As you can see, parameters inside <fields> section can be a key-value pair or just a key (a name). If a parameter is a key-value, the key will be the name of the field inside the "fields" JSON object, whereas the value is the field name of the input event. So a key-value pair is a rename.

If a parameter has just a key, it means its value is exactly the same as the key.

When data_type is metric

For metrics, parameters inside <fields> are used as dimensions. If <fields> is not presented, the original input event will be used as dimensions. If an empty <fields></fields> is presented, no dimension will be sent. For example, given configuration like

<match **>
  @type splunk_hec
  data_type metric
  ...omitting other parameters...

  metric_name_key name
  metric_value_key value
  <fields>
    file
    level
    app applicatioin
  </fields>
</match>

and an input event like

{"application": "webServer", "file": "server.rb", "value": 100, "status": "OK", "message": "Normal", "name": "CPU Usage"}

Then, a metric of "CPU Usage" with value=100, along with 3 dimensions file="server.rb", status="OK", and app="webServer" will be sent to Splunk.

<format> section (optional) (multiple)

The <format> section let us define which formatter to use to format events. By default, it uses the json formatter.

Besides the @type parameter, you should define all other parameters for the formatter inside this section.

Multiple <format> sections can be defined to use different formatters for different tags. Each <format> section accepts an argument just like the <match> section does, to define tag matching. But default, every event will be formatted with json. For example:

<match **>
  @type splunk_hec
  ...

  <format sometag.**>
    @type single_value
    message_key log
  </format>

  <format some.othertag>
    @type csv
    fields ["some", "fields"]
  </format>

In this example, it will format events with tags which start with sometag. with the single_value formatter, and format events with tags some.othertag with the csv formatter, and format all other events with the json formatter (the default formatter).

If you want to use a different default formatter, you can add a <format **> (or <format>) section.

@type (string) (required)

Defines which formatter to use.

Net::HTTP::Persistent parameters (optional)

The following parameters can be used for tuning HTTP connections

idle_timeout (integer)

The default is 5 seconds. If a connection has not been used for this number of seconds it will automatically be reset upon the next use to avoid attempting to send to a closed connection; nil means no timeout.

read_timeout (integer)

The default is nil. The amount of time allowed between reading two chunks from the socket.

open_timeout (integer)

The default is nil. The amount of time to wait for a connection to be opened.

SSL parameters

There are quite some parameters you can use to configure SSL (for HTTPS protocol). All these parameters are optional.

client_cert (string)

The path to a file containing a PEM-format CA certificate for this client.

client_key (string)

The private key for this client.

ca_file (string)

The path to a file containing a PEM-format CA certificate.

ca_path (string)

The path to a directory containing CA certificates in PEM format.

ciphers (array)

List of SSl ciphers allowed.

insecure_ssl (bool)

Indicates if insecure SSL connection is allowed, i.e. do not verify the server's certificate.

Default value: false.

About Buffer

This plugin sends events to HEC using batch mode. It batches all events in a chunk in one request. So you need to configure the <buffer> section carefully to gain the best performance. Here are some hints:

  • Read through the fluentd buffer document to understand the buffer configurations.
  • Use chunk_limit_size and/or chunk_limit_records to define how big a chunk can be. And remember that all events in a chunk will be sent in one request.
  • Splunk has a limit on how big the payload of a HEC request can be. And it's defined with max_content_length in the [http_input] section of limits.conf. In Splunk of version 6.5.0+, the default value is 800MiB, while in versions before 6.5.0, it's just 1MB. Make sure your chunk size won't exceed this limit, or you should change the limit on your Splunk deployment.
  • Sending requests to HEC takes time, so if you flush your fluentd buffer too fast (for example, with a very small flush_interval), it's possible that the plugin cannot catch up with the buffer flushing. There are two ways you can handle this situation, one is to increase the flush_interval or use multiple flush threads by setting flush_thread_count to a number bigger than 1.

License

Please see LICENSE.