This project provides two OpenTSDB(v2.1.3) rpc plugins.

The Thrift Plugin

This is just a simple Thrift server that will be start when OpenTSDB starts.

By default, it binds to 0.0.0.0, which is configable var property em.thrift.host and listen port is 9999, which also can be changed using property em.thrift.port

The thrift IDL used here can be found at src/main/thrift/metrics.thrift which will provided very detailed view of how to use this thrift service.

As defined in the IDL, this server accepts two types of metrics data, simple text or Tsdata which is a predefined thrift struct.

The plain text type data is just like the OpenTSDB's telent service, so one can write a thrift client and put same data that used in the telnet service, it uses same metrics syntax.

The Kafka Plugin

While, this is somehow misusing the OpenTSDB's RPC mechanism, since it is not a RPC plugin at all.

What this plugin does is simple start a kafka consumer and fetch data from kafka and then write it to OpenTSDB.

This kafka consumer reads one more topics, you can shot data only type of plain text or tsdata's binary into every topic.

Note: the topic name property has no default value, so if there is no such property provided, then this plugin will not fetch any data from kafka but just quite.

We can get you topics via this configured property em.kafka.topics, you must split them by ,, they are just index, real topic name will be provode by the property named em.kafka.topicN.name e.g: em.kafka.topics=topic1,topic2,...,topicN

For each topic there are some properties that should be defined in the configuration file: (e.g: topic1)

1. em.kafka.topic1.name

   topic names of kafka

2. em.kafka.topic1.datatype

   the data type of your data shot into this topic, only two value for you : plaintext for plain text value and tsdata for tsdata's binary. Metrics schemas:

   • plaintext (just only one space between fragements):
     • (name tags timestamp value)
     • sys.cpu.user host=webserver01,cpu=0 1356998400 1
   • tsdata:
     • TSData(name:sys.cpu.user, value:6.0, timestamp:1356998400, tags:{host=webserver01, cpu=0})

3. em.kafka.topic1.zookeeper.connect

   The zk host which will be used for the kafka consumer to connect to.

4. em.kafka.topic1.group.id

   The client id used to identify kafka consumer group.

How to build

    ./gradlew clean build assembleMainDist

The artifact will be build and save to build/distributions/

NOTE: this requires Thrift 0.9.3

Install

Copy the build artifact and copy the jars to the OpenTSDB's lib dir, please note some dependencies of this project maybe already exist in the OpenTSDB's distribution, so just skip such jars and only copy non-exist ones.

Then modify OpenTSDB's conf file and add properties required by this plugin, all properties can be found in the class cn.kenshinn.tsdb.thrift.rpc.Constants

To enable the rpc plugins provided by this project, please make sure such line exist in the Opentsdb's conf

tsd.rpc.plugins=cn.kenshinn.thrift.tsdb.rpc.ThriftTSDBRpcPlugin,cn.kenshinn.tsdb.kafka.plugin.KafkaConsumerRPCPlugin