Supported file formats

To take advantage of Apache Druid's querying capabilities, you ingest or load your original data from an input source. This topic is a reference for the supported input formats and sources.

Natively supported input formats by type

Druid supports the following input file types natively without loading an extension:

JSON

• Streaming ingestion: Kafka, Kinesis
  
• Batch ingestion: Index parallel (index), Azure blob
  Druid supports Newline delimited JSON. See also JSON data format and Data ingestion tutorial.

Comma separated values (CSV)

• Streaming ingestion: Kafka, Kinesis
  
• Batch ingestion: Index parallel (index)
  See CSV data format.

Tab separated values (TSV) and custom delimiters

• Streaming ingestion: Kafka, Kinesis
  
• Batch ingestion: Index parallel (index)
  See TSV data format.

Input formats supported with an extension by type

Druid supports the following input file types through an extension:

Avro OCF

• Extension: druid-avro-extensions
  
• Batch ingestion: Index parallel (index), or Hadoop with the Avro Hadoop parser
  See Avro OCF data format.

Avro Stream

• Extension: druid-avro-extensions
  
• Streaming ingestion: Kafka, Kinesis See Avro stream data format.

Parquet

• Extension: druid-parquet-extensions
  
• Streaming ingestion: Kafka, Kinesis
  
• *Batch ingestion: Index parallel (index), or Hadoop with the Parquet Hadoop parser
  See Parquet data format.

Orc

• Extension: druid-orc-extensions
  
• Batch ingestion: Index parallel (index), or Hadoop with the Orc Hadoop parser
  See Orc data format.

Protobuf

• Extension: druid-protobuf-extensions
  
• Streaming ingestion: Kafka, Kinesis
  See Protobuf parser.

Supported input formats by ingestion type

This section lists the file formats Druid supports for each ingestion type.

Streaming ingestion

Druid supports the following file formats for streaming ingestion:

Kafka

• JSON
• CSV
• TSV
• Avro stream
• Protobuf

Kinesis

• JSON
• CSV
• TSV and custom delimiter
• Avro stream
• Protobuf

Azure blob

• JSON

Batch ingestion

Druid supports the following file formats for batch ingestion:

Index parallel (index)

• JSON
• CSV
• TSV and custom delimiter
• Parquet
• Avro

Hadoop batch ingestion

• Avro
• Parquet
• Orc

Learn more

See the following topics for more information:

• Ingestion overview for basic information about ingestion.
• Data formats for examples and configuration options.