Ingestion Job Spec

The ingestion job spec is used while generating, running, and pushing segments from the input files.

The Job spec can be in either YAML or JSON format (0.5.0 onwards). Property names remain the same in both formats.

To use the JSON format, add the propertyjob-spec-format=jsonin the properties file while launching the ingestion job. The properties file can be passed as follows

pinot-admin.sh LaunchDataIngestionJob \
-jobSpecFile /path/to/job_spec.json \
-propertyFile /path/to/job.properties

Template your job spec file

Users are allowed to define some variables in the job spec file to make it a template then passing the variables at runtime.

Templating is based on Groovy SimpleTemplateEngine.

E.g. users can specify below in the job spec file:

inputDirURI: 'file:///path/to/input/${year}/${month}/${day}/${hour}'

The values for the template strings in the jobSpecFile can be passed in one of the following three ways mentioned in their order of precedence, for same key, 1 will override 2 will override 3.

1. Values from the -values array passed from the Cmd Line. See Launch Data Ingestion Job

2. Values from the environment variables

3. Values from the propertyFile

Still take above inputDirURI as example,

We can define a job.config file with below content:

year=2020
month=05
day=01
hour=00

Above properties can be override by environment variables

export month=06
export day=02
export hour=03

From the command line, user can further override those keys using flag -values,

pinot-admin.sh LaunchDataIngestionJob \
-jobSpecFile /path/to/job_spec.json \
-propertyFile job.confg \
-values day=03 hour=04

After that the real ingestion spec passed to ingestion job will have inputDirURI as 'file:///path/to/input/2020/06/03/04'

Ingestion Job Spec

The following configurations are supported by Pinot

Top-Level Spec

Property	Description
executionFrameworkSpec	Contains config related to the executor to use to ingest data. See Execution Framework Spec
jobType	Type of job to execute. The following types are supported SegmentCreation SegmentMetadataPush SegmentTarPush SegmentUriPush SegmentCreationAndMetadataPush: (Recommended for production environments where Pinot deep store is configured): Use this job to bypass the controller, and send the segment payload directly to the data store. SegmentCreationAndUriPush: (Alternative option if Pinot deep store is configured) Use this job to create the segment on the deep store, push the URI to the controller to download the segment, extract metadata from the URI, and copy the data to deep store. SegmentCreationAndTarPush: If you use Network File System (NFS) or something that sits behind a controller, and are unable to externally copy segments to the data store, use this job to push the segment payload. Note: For production environments where Pinot Deep Store is configured, it's recommended to use SegmentCreationAndMetadataPush
inputDirURI	Absolute Path along with scheme of the directory containing all the files to be ingested, e.g. s3://bucket/path/to/input, /path/to/local/input
includeFileNamePattern	Only Files matching this pattern will be included from inputDirURI. Both glob and regex patterns are supported. Examples: Use 'glob:.avro'or 'regex:^..(avro)$' to include all avro files one level deep in the inputDirURI. Alternatively, use 'glob:*/.avro' to include all the avro files in inputDirURI as well as its subdirectories - bear in mind that, with this approach, the pattern needs to match the absolute path. You can use Glob tool or Regex Tool to test out your patterns.
excludeFileNamePattern	Exclude file name pattern, supported glob pattern. Similar usage as includeFilePatternName
searchRecursively	Set to true to explicitly search input files recursively from inputDirURI. It is set to true by default for now.
outputDirURI	Absolute Path along with scheme of the directory where to output all the segments.
overwriteOutput	Set to true to overwrite segments if already present in the output directory. Or set tofalseto raise exceptions.
pinotFSSpecs	List of all the filesystems to be used for ingestions. You can mention multiple values in case input and output directories are present in different filesystems. For more details, scroll down to Pinot FS Spec.
tableSpec	Defines table name and where to fetch corresponding table config and table schema. For more details, scroll down to Table Spec.
recordReaderSpec	Parser to use to read and decode input data. For more details, scroll down to Record Reader Spec.
segmentNameGeneratorSpec	Defines how the names of the segments will be. For more details, scroll down to Segment Name Generator Spec.
pinotClusterSpecs	Defines the Pinot Cluster Access Point. For more details, scroll down to Pinot Cluster Spec.
pushJobSpec	Defines segment push job-related configuration. For more details, scroll down to Push Job Spec.

Example

executionFrameworkSpec:
  name: 'spark'
  segmentGenerationJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.spark.SparkSegmentGenerationJobRunner'
  segmentTarPushJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.spark.SparkSegmentTarPushJobRunner'
  segmentUriPushJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.spark.SparkSegmentUriPushJobRunner'
  segmentMetadataPushJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.spark.SparkSegmentMetadataPushJobRunner'
  extraConfigs:
    stagingDir: hdfs://examples/batch/airlineStats/staging

# Recommended to set jobType to SegmentCreationAndMetadataPush for production environments where Pinot Deep Store is configured
jobType: SegmentCreationAndTarPush

inputDirURI: 'examples/batch/airlineStats/rawdata'
includeFileNamePattern: 'glob:**/*.avro'
searchRecursively: true
outputDirURI: 'hdfs:///examples/batch/airlineStats/segments'
overwriteOutput: true
pinotFSSpecs:
  - scheme: hdfs
    className: org.apache.pinot.plugin.filesystem.HadoopPinotFS
  - scheme: file
    className: org.apache.pinot.spi.filesystem.LocalPinotFS
recordReaderSpec:
  className: 'org.apache.pinot.plugin.inputformat.avro.AvroRecordReader'
tableSpec:
  tableName: 'airlineStats'
  schemaURI: 'http://localhost:9000/tables/airlineStats/schema'
  tableConfigURI: 'http://localhost:9000/tables/airlineStats'
segmentNameGeneratorSpec:
  type: normalizedDate
  configs:
    segment.name.prefix: 'airlineStats_batch'
    exclude.sequence.id: true
pinotClusterSpecs:
  - controllerURI: 'http://localhost:9000'
pushJobSpec:
  pushParallelism: 2
  pushAttempts: 2
  pushRetryIntervalMillis: 1000

Execution Framework Spec

The configs specify the execution framework to use to ingest data. Check out Batch Ingestion for configs related to all the supported frameworks

Property	Description
name	name of the execution framework. can be one of spark,hadoop or standalone
segmentGenerationJobRunnerClassName	The class name implements org.apache.pinot.spi.ingestion.batch.runner.IngestionJobRunner interface to run the segment generation job
segmentTarPushJobRunnerClassName	The class name implements org.apache.pinot.spi.ingestion.batch.runner.IngestionJobRunner interface to push the segment TAR file
segmentUriPushJobRunnerClassName	The class name implements org.apache.pinot.spi.ingestion.batch.runner.IngestionJobRunner interface to send segment URI
segmentMetadataPushJobRunnerClassName	The class name implements org.apache.pinot.spi.ingestion.batch.runner.IngestionJobRunner interface to send segment Metadata
extraConfigs	Key-value pairs of configs related to the framework of the executions

Example

executionFrameworkSpec:
    name: 'standalone'
    segmentGenerationJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.standalone.SegmentGenerationJobRunner'
    segmentTarPushJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.standalone.SegmentTarPushJobRunner'
    segmentUriPushJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.standalone.SegmentUriPushJobRunner'
    segmentMetadataPushJobRunnerClassName: 'org.apache.pinot.plugin.ingestion.batch.standalone.SegmentMetadataPushJobRunner'

Pinot FS Spec

field	description
schema	used to identify a PinotFS. E.g. local, hdfs, dbfs, etc
className	Class name used to create the PinotFS instance. E.g. org.apache.pinot.spi.filesystem.LocalPinotFS is used for local filesystem org.apache.pinot.plugin.filesystem.HadoopPinotFS is used for HDFS
configs	configs used to init PinotFS instance

Table Spec

Table spec is used to specify the table in which data should be populated along with schema.

Property	Description
tableName	name of the table in which to populate the data
schemaURI	location from which to read the schema for the table. Supports both File systems as well as HTTP URI
tableConfigURI	location from which to read the config for the table. Supports both File systems as well as HTTP URI

Example

tableSpec:
  tableName: 'airlineStats'
  schemaURI: 'http://localhost:9000/tables/airlineStats/schema'
  tableConfigURI: 'http://localhost:9000/tables/airlineStats'
​

Record Reader Spec

field	description
dataFormat	Record data format, e.g. 'avro', 'parquet', 'orc', 'csv', 'json', 'thrift' etc.
className	Corresponding RecordReader class name. E.g. org.apache.pinot.plugin.inputformat.avro.AvroRecordReader org.apache.pinot.plugin.inputformat.csv.CSVRecordReader org.apache.pinot.plugin.inputformat.parquet.ParquetRecordReader org.apache.pinot.plugin.inputformat.json.JSONRecordReader org.apache.pinot.plugin.inputformat.orc.ORCRecordReader org.apache.pinot.plugin.inputformat.thrift.ThriftRecordReader
configClassName	Corresponding RecordReaderConfig class name, it's mandatory for CSV and Thrift file format. E.g. org.apache.pinot.plugin.inputformat.csv.CSVRecordReaderConfig org.apache.pinot.plugin.inputformat.thrift.ThriftRecordReaderConfig
configs	Used to init RecordReaderConfig class name, this config is required for CSV and Thrift data format.

Segment Name Generator Spec

Property	Description
type	The type of name generator to use. If not specified, an appropriate type will be inferred based on segment generator config properties. The following values are supported - simple - this is the default spec. normalizedDate - use this type when the time column in your data is in the String format instead of epoch time. fixed - configure the segment name by the user. inputFile - supports naming the resulting segment file based on the input file name & path. Use this if your table doesn't have a time column. Ensure that input file names are unique though otherwise will lead to several issues.
configs	Configs to init SegmentNameGenerator
segment.name	For fixed SegmentNameGenerator. Explicitly set the segment name.
segment.name.postfix	For simple SegmentNameGenerator. Postfix will be appended to all the segment names.
segment.name.prefix	For normalizedDate SegmentNameGenerator. The Prefix will be prepended to all the segment names.
exclude.sequence.id	Whether to include sequence ids in segment name. Needed when there are multiple segments for the same time range.
use.global.directory.sequence.id	Assign sequence ids to input files based on all input files under the directory. Set to false to use local directory sequence id. This is useful when generating multiple segments for multiple days. In that case, each of the days will start from sequence id 0.
append.uuid.to.segment.name	If the input data doesn't contain a time column, set this to true to generate unique segment names. Can be used with any name generator type.
file.path.pattern	For inputFile, a Java regular expression used to match against the input file URI. e.g. '.+/(.+).gz' to extract file name from a .gz file without the extension
segment.name.template	For inputFile , the string template that should be used to substitute extracted fileName. Currently only supports ${filePathPattern:<match group>}

Example

segmentNameGeneratorSpec:
  type: normalizedDate
  configs:
    segment.name.prefix: 'airlineStats_batch'
    exclude.sequence.id: true

To set the segment name to be the same as the input file name (without the trailing .gz), use:

segmentNameGeneratorSpec:
  type: inputFile
  configs:
    file.path.pattern: '.+/(.+)\.gz'
    segment.name.template: '\${filePathPattern:\1}'

Note that $ in the yaml file must be escaped, since Pinot uses Groovy's SimpleTemplateEngine to process the yaml file, and a raw $ is treated as a template specifier.

Pinot Cluster Spec

Property	Description
controllerURI	URI to use to fetch table/schema information and push data

Example

pinotClusterSpecs:
  - controllerURI: 'http://localhost:9000'

Push Job Spec

Property	Description
pushAttempts	Number of attempts for push job. Default is 1, which means no retry
pushParallelism	Workers to use for push job. Default is 1
pushRetryIntervalMillis	Time in milliseconds to wait for between retry attempts Default is 1 second.
segmentUriPrefix	append this string before the path of the push destination. Generally, it is the scheme of the filesystem e.g. s3:// , file:// etc.
segmentUriSuffix	append this string after the path of the push destination.
pushFileNamePattern	segment name pattern for which segments to push, supported glob and regex patterns. E.g. 'glob:**2023-01*' will push all the segment files under the outputDirURI whose names contain '2023-01'.
batchSegmentUpload	Boolean field for which the default value is false. When the value is set to true segments are uploaded in batch mode which is faster than uploading segments one after the other. Works when the jobType is set to SegmentMetadataPush

Example

pushJobSpec:
  pushParallelism: 2
  pushAttempts: 2
  pushRetryIntervalMillis: 1000
  segmentUriPrefix : 'file://'
  segmentUriSuffix : my-dir/
  pushFileNamePattern : glob:\*\*2023-01\*