datasource kafka: Add in_kafka plugin description

content/datasources/kafka.md

@@ -11,7 +11,32 @@ Fluentd can setup to collect messages from Kafka. Applications include:
 1. Sending Kafka messages into HDFS for analysis
 2. Sending Kafka messages into Elasticsearch for analysis
 
-## Setup
+You can two choices for this purpose whether using `in_kafka` or using `kafka-fluentd-consumer`.
+
+## Setup: fluent-plugin-kafka
+
+1. Install the [Kafka input plugin](https://github.com/htgc/fluent-plugin-kafka) by running the following command:
+
+    ```
+    $ fluent-gem install fluent-plugin-kafka
+    ```
+
+2. Open your Fluentd configuration file and add the following lines:
+
+    ```
+    <source>
+      @type  kafka
+      host   <broker host>
+      port   <broker port: default=9092>
+      topics <listening topics(separate with comma',')>
+      format <input text type (text|json|ltsv|msgpack)>
+      message_key <key (Optional, for text format only, default is message)>
+    </source>
+    ```
+
+    With the above setup, Fluentd consumes Kafka messages via `in_kafka` plugin.
+
+## Setup: kafka-fluentd-consumer
 
 1. Download the latest [kafka-fluentd-consumer jar](https://github.com/treasure-data/kafka-fluentd-consumer/releases).
 
@@ -29,3 +54,8 @@ Fluentd can setup to collect messages from Kafka. Applications include:
     ```
 
    With the above setup, Fluentd consumes Kafka messages which are specified topics in `fluentd-consumer.properties` via `in_exec` plugin.
+
+### Note
+
+For simplification, you can use `in_kafka` plugin to retrive kafka messages.
+If you assume highly kafka traffic in production, we recommend to use `kafka-fluentd-consumer` instead of `in_kafka`. Because `in_kafka` has been reported high CPU usage when 1000req/sec environment. In more detail, please refer to [the issue](https://github.com/htgc/fluent-plugin-kafka/issues/16).