Storing Kafka Data in Hive Through Flume

Last updated: 2020-02-28 20:02:32


Scenario Description

Data in Kafka can be collected through Flume and stored in Hive.

Preparations for Development

  • As this job requires access to CKafka, you need to create a CKafka instance first. For more information, please see CKafka.
  • Confirm that you have activated Tencent Cloud and created an EMR cluster. When creating the EMR cluster, you need to select the Spark component on the software configuration page.

Using the Kafka Toolkit in the EMR Cluster

First, you need to check the private IP and port number of CKafka. Log in to the CKafka Console, select the CKafka instance you want to use, and view its private IP as $kafkaIP in the basic information section, and the port number is generally 9092 by default. Create a topic named kafka_test on the topic management page.

Configuring Flume

  1. Create the Flume configuration file
    agent.sources = kafka_source
    agent.channels = mem_channel
    agent.sinks = hive_sink
    # The following code is used to configure the source
    agent.sources.kafka_source.type = org.apache.flume.source.kafka.KafkaSource
    agent.sources.kafka_source.channels = mem_channel
    agent.sources.kafka_source.batchSize = 5000
    agent.sources.kafka_source.kafka.bootstrap.servers = $kafkaIP:9092
    agent.sources.kafka_source.kafka.topics = kafka_test
    # The following code is used to configure the sink = mem_channel
    agent.sinks.hive_sink.type = hive
    agent.sinks.hive_sink.hive.metastore = thrift://
    agent.sinks.hive_sink.hive.database = default
    agent.sinks.hive_sink.hive.table = weblogs
    agent.sinks.hive_sink.hive.partition = asia,india,%y-%m-%d-%H-%M
    agent.sinks.hive_sink.useLocalTimeStamp = true
    agent.sinks.hive_sink.round = true
    agent.sinks.hive_sink.roundValue = 10
    agent.sinks.hive_sink.roundUnit = minute
    agent.sinks.hive_sink.serializer = DELIMITED
    agent.sinks.hive_sink.serializer.delimiter = ","
    agent.sinks.hive_sink.serializer.serdeSeparator = ','
    agent.sinks.hive_sink.serializer.fieldnames =id,msg
    # The following code is used to configure the channel
    agent.channels.mem_channel.type = memory
    agent.channels.mem_channel.capacity = 100000
    agent.channels.mem_channel.transactionCapacity = 10000
    You can confirm Hive Metastore in the following way:
    grep "hive.metastore.uris" -C 2 /usr/local/service/hive/conf/hive-site.xml
  2. Create a Hive table.
    create table weblogs ( id int , msg string )
    partitioned by (continent string, country string, time string)
    clustered by (id) into 5 buckets
    stored as orc TBLPROPERTIES ('transactional'='true');

    All the following conditions must be met: it must be a table with partitions and buckets, the storage format is ORC, and TBLPROPERTIES ('transactional'='true') is set.

  3. Enable the Hive transaction.
    In the console, add the following configuration items to hive-site.xml.

    After the configuration is distributed and Hive is restarted, the hadoop-hive log will prompt that the Metastore cannot be connected to. Please ignore this error. Because of the startup order of the processes, Metastore will be started before HiveServer2.

  4. Copy hive-hcatalog-streaming-xxx.jar of Hive to the lib directory of Flume.
    cp -ra /usr/local/service/hive/hcatalog/share/hcatalog/hive-hcatalog-streaming-2.3.3.jar /usr/local/service/flume/lib/
  5. Run Flume.
    ./bin/flume-ng agent --conf ./conf/ -f -n agent -Dflume.root.logger=INFO,console
  6. Run the Kafka producer.
    [hadoop@172 kafka]$ ./bin/ --broker-list $kafkaIP:9092 --topic kafka_test


  • Enter information on the client of the Kafka producer and press Enter.
  • Check whether there is corresponding data in the Hive table.

Reference Documentation