Configuration Reference for HDFS 2 Sink Connector for Confluent Platform¶
To use this connector, specify the name of the connector class in the
connector.class
configuration property.
connector.class=io.confluent.connect.hdfs.HdfsSinkConnector
Connector-specific configuration properties are described below.
HDFS¶
hdfs.url
The HDFS connection URL. This configuration has the format of hdfs://hostname:port and specifies the HDFS to export data to. This property is deprecated and will be removed in future releases. Use
store.url
instead.- Type: string
- Default: null
- Importance: high
hadoop.conf.dir
The Hadoop configuration directory.
- Type: string
- Default: “”
- Importance: high
hadoop.home
The Hadoop home directory.
- Type: string
- Default: “”
- Importance: high
logs.dir
Top level directory to store the write ahead logs. Supports
${topic}
in the value, which will be replaced by the actual topic name. Supports${0}
, …,${n}
in conjunction withtopic.capture.groups.regex
. For details, seetopic.capture.groups.regex
configuration documentation.- Type: string
- Default: logs
- Importance: high
Security¶
hdfs.authentication.kerberos
Configuration indicating whether HDFS is using Kerberos for authentication.
- Type: boolean
- Default: false
- Importance: high
- Dependents:
connect.hdfs.principal
,connect.hdfs.keytab
,hdfs.namenode.principal
,kerberos.ticket.renew.period.ms
connect.hdfs.principal
The principal name to load from the keytab for Kerberos authentication.
- Type: string
- Default: “”
- Importance: high
connect.hdfs.keytab
The path to the keytab file for the HDFS connector principal. This keytab file should only be readable by the connector user.
- Type: string
- Default: “”
- Importance: high
hdfs.namenode.principal
The Kerberos principal name for the HDFS Namenode. The format is
nn/_HOST@REALM.TLD
.- Type: string
- Default: “”
- Importance: high
kerberos.ticket.renew.period.ms
The period in milliseconds to renew the Kerberos ticket.
- Type: long
- Default: 3600000 (1 hour)
- Importance: low
Connector¶
format.class
The format class to use when writing data to the store. Format classes implement the
io.confluent.connect.storage.format.Format
interface.- Type: class
- Default:
io.confluent.connect.hdfs.avro.AvroFormat
- Importance: high
These classes are available by default:
io.confluent.connect.hdfs.avro.AvroFormat
io.confluent.connect.hdfs.json.JsonFormat
io.confluent.connect.hdfs.orc.OrcFormat
io.confluent.connect.hdfs.parquet.ParquetFormat
io.confluent.connect.hdfs.string.StringFormat
flush.size
Number of records written to store before invoking file commits.
- Type: int
- Importance: high
Important
Rotation strategy logic: In the following rotation strategies, the logic to flush files to storage is triggered when a new record arrives, after the defined interval or scheduled interval time. Flushing files is also triggered periodically by the
offset.flush.interval.ms
setting defined in the Connect worker configuration. Theoffset.flush.interval.ms
setting defaults to 60000 ms (60 seconds). If you enable the propertiesrotate.interval.ms
orrotate.schedule.interval.ms
and ingestion rate is low, you should setoffset.flush.interval.ms
to a smaller value so that records flush at the rotation interval (or close to the interval) . Leaving theoffset.flush.interval.ms
set to the default 60 seconds may cause records to stay in an open file for longer than expected, if no new records get processed that trigger rotation.rotate.interval.ms
The time interval in milliseconds to invoke file commits. This configuration ensures that file commits are invoked every configured interval. This configuration is useful when data ingestion rate is low and the connector didn’t write enough messages to commit files. The default value -1 means that this feature is disabled.
- Type: long
- Default: -1
- Importance: high
rotate.schedule.interval.ms
The time interval in milliseconds to periodically invoke file commits. This configuration ensures that file commits are invoked every configured interval. Time of commit will be adjusted to 00:00 of selected timezone. Commit will be performed at scheduled time regardless previous commit time or number of messages. This configuration is useful when you have to commit your data based on current server time, like at the beginning of every hour. The default value -1 means that this feature is disabled.
- Type: long
- Default: -1
- Importance: medium
The following Avro converter properties can be used in the connector configuration:
schemas.cache.config
The size of the schema cache used in the Avro converter.
- Type: int
- Default: 1000
- Importance: low
enhanced.avro.schema.support
Enable enhanced Avro schema support in the Avro Converter. When set to
true
, this property preserves Avro schema package information and Enums when going from Avro schema to Connect schema. This information is added back in when going from Connect schema to Avro schema.- Type: boolean
- Default: true
- Importance: low
connect.meta.data
Allow the Connect converter to add its metadata to the output schema.
- Type: boolean
- Default: true
- Importance: low
The
connect.meta.data
property preserves the following Connect schema metadata when going from Connect schema to Avro schema. The following metadata is added back in when going from Avro schema to Connect schema.- doc
- version
- parameters
- default value
- name
- type
For detailed information and configuration examples for Avro converters listed above, see Using Kafka Connect with Schema Registry.
retry.backoff.ms
The retry backoff in milliseconds. This config is used to notify Connect to retry delivering a message batch or performing recovery in case of transient exceptions.
- Type: long
- Default: 5000 (5 seconds)
- Importance: low
shutdown.timeout.ms
Clean shutdown timeout. This makes sure that asynchronous Hive metastore updates are completed during connector shutdown.
- Type: long
- Default: 3000 (3 seconds)
- Importance: medium
filename.offset.zero.pad.width
Width to zero-pad offsets in store’s filenames if offsets are too short in order to provide fixed-width filenames that can be ordered by simple lexicographic sorting.
- Type: int
- Default: 10
- Valid Values: [0,…]
- Importance: low
avro.codec
The Avro compression codec to be used for output files. Available values: null, deflate, snappy and bzip2 (Codec source is org.apache.avro.file.CodecFactory).
- Type: string
- Default: null
- Valid Values: [null, deflate, snappy, bzip2]
- Importance: low
parquet.codec
The Parquet compression codec to be used for output files.
- Type: string
- Default: snappy
- Valid Values: [none, snappy, gzip, brotli, lz4, lzo, zstd]
- Importance: low
Storage¶
storage.class
The underlying storage layer.
- Type: class
- Default: io.confluent.connect.hdfs.storage.HdfsStorage
- Importance: high
topics.dir
Top level directory to store the data ingested from Apache Kafka®. Supports
${topic}
in the value, which will be replaced by the actual topic name. Supports${0}
, …,${n}
in conjunction withtopic.capture.groups.regex
. For details, seetopic.capture.groups.regex
configuration documentation.- Type: string
- Default: topics
- Importance: high
store.url
Store’s connection URL, if applicable. For example:
hdfs://hostname:port
.- Type: string
- Default: null
- Importance: high
directory.delim
Directory delimiter pattern
- Type: string
- Default: /
- Importance: medium
topic.capture.groups.regex
A Java Pattern regex that matches the entire topic and captures values for substituting into
topics.dir
. Indexed capture groups are accessible with${n}
, where${0}
refers to the whole match and${1}
refers to the first capture group. Example config value of([a-zA-Z]*)_([a-zA-Z]*)
will match topics that are two words delimited by an underscore and will capture each word separately. Withtopic.dir = ${1}/${2}
, a record from the topicexample_name
will be written into a subdirectory ofexample/name/
. By default, this functionality is not enabled.- Type: string
- Default: null
- Importance: low
Partitioner¶
partitioner.class
The partitioner to use when writing data to the store. The following partitioners are available:
DefaultPartitioner
preserves the Kafka partitions.DailyPartitioner
partitions data according to date.HourlyPartitioner
partitions data according to hour.FieldPartitioner
partitions the data to different directories according to the value of the partitioning field specified inpartition.field.name
.TimeBasedPartitioner
partitions data according to ingestion time.
- Type: class
- Default: io.confluent.connect.storage.partitioner.DefaultPartitioner
- Importance: high
- Dependents:
partition.field.name
,partition.duration.ms
,path.format
,locale
,timezone
partition.field.name
The name of the partitioning field when
FieldPartitioner
is used. You can enter multiple partitioning field names using comma-separated names.- Type: list
- Default: “”
- Importance: medium
partition.duration.ms
The duration of a partition milliseconds used by
TimeBasedPartitioner
. The default value -1 means that we are not usingTimeBasedPartitioner
.- Type: long
- Default: -1
- Importance: medium
path.format
This configuration is used to set the format of the data directories when partitioning with
TimeBasedPartitioner
. The format set in this configuration converts the Unix timestamp to proper directories strings. For example, if you setpath.format='year'=YYYY/'month'=MM/'day'=dd/'hour'=HH
, the data directories will have the format/year=2015/month=12/day=07/hour=15/
.- Type: string
- Default: “”
- Importance: medium
locale
The locale to use when partitioning with
TimeBasedPartitioner
. Used to format dates and times. For example, useen-US
for US English,en-GB
for UK English, orfr-FR
for French (in France). These may vary by Java version. See the available locales.- Type: string
- Default: “”
- Importance: medium
timezone
The timezone to use when partitioning with
TimeBasedPartitioner
. Used to format and compute dates and times. All timezone IDs must be specified in the long format, such asAmerica/Los_Angeles
,America/New_York
, andEurope/Paris
, orUTC
. Alternatively a locale independent, fixed offset, datetime zone can be specified in form[+-]hh:mm
. Support for these timezones may vary by Java version. See the available timezones within each locale, such as those within the US English locale.- Type: string
- Default: “”
- Valid Values: Any timezone accepted by: class
org.joda.time.DateTimeZone
- Importance: medium
timestamp.extractor
The extractor that gets the timestamp for records when partitioning with
TimeBasedPartitioner
. It can be set toWallclock
,Record
orRecordField
in order to use one of the built-in timestamp extractors or be given the fully-qualified class name of a user-defined class that extends theTimestampExtractor
interface.- Type: string
- Default: Wallclock
- Importance: medium
timestamp.field
The record field to be used as timestamp by the timestamp extractor.
- Type: string
- Default: timestamp
- Importance: medium
Hive¶
hive.integration
Configuration indicating whether to integrate with Hive when running the connector.
- Type: boolean
- Default: false
- Importance: high
- Dependents:
hive.metastore.uris
,hive.conf.dir
,hive.home
,hive.database
hive.metastore.uris
The Hive metastore URIs, can be IP address or fully-qualified domain name and port of the metastore host.
- Type: string
- Default: “”
- Importance: high
hive.conf.dir
Hive configuration directory
- Type: string
- Default: “”
- Importance: high
hive.home
Hive home directory.
- Type: string
- Default: “”
- Importance: high
hive.database
The database to use when the connector creates tables in Hive.
- Type: string
- Default: default
- Importance: high
hive.table.name
The hive table name to use. It must contain
${topic}
to inject the corresponding topic name.- Type: string
- Default: “${topic}”
- Importance: low
Schema¶
schema.compatibility
The schema compatibility rule to use when the connector is observing schema changes. The supported configurations are NONE, BACKWARD, FORWARD and FULL.
- Type: string
- Default: NONE
- Importance: high