Amazon S3 Sink Connector for Confluent Platform

The Amazon S3 Sink connector exports data from Apache Kafka® topics to S3 objects in either Avro, JSON, or Bytes formats. Depending on your environment, the S3 connector can export data by guaranteeing exactly-once delivery semantics to consumers of the S3 objects it produces.

The Amazon S3 sink connector periodically polls data from Kafka and in turn uploads it to S3. A partitioner is used to split the data of every Kafka partition into chunks. Each chunk of data is represented as an S3 object. The key name encodes the topic, the Kafka partition, and the start offset of this data chunk. If no partitioner is specified in the configuration, the default partitioner which preserves Kafka partitioning is used. The size of each data chunk is determined by the number of records written to S3 and by schema compatibility.

Features

The Amazon S3 Sink connector for Confluent Platform includes the following features:

• Exactly-once delivery
• Dead Letter Queue
• Multiple tasks
• Pluggable data format with or without schema
• Pluggable partitioner
• Non-AWS object storage support
• Writing record keys and headers

Exactly-once delivery

Records that are exported using a deterministic partitioner are delivered with exactly-once semantics regardless of the eventual consistency of S3.

Note that if versioning is enabled for the S3 bucket, you might see multiple versions of the same file in S3; but, if you view the most recent version among those files, you will see that the persistence of data exactly once remains valid.

Dead Letter Queue

This connector supports the Dead Letter Queue (DLQ) functionality. For information about accessing and using the DLQ, see Confluent Platform Dead Letter Queue.

Multiple tasks

The Amazon S3 Sink connector supports running one or more tasks. You can specify the number of tasks in the tasks.max configuration parameter. This can lead to performance gains when multiple files need to be parsed.

Pluggable data format with or without schema

Out of the box, the connector supports writing data to S3 in Avro and JSON format. Besides records with schema, the connector supports exporting plain JSON records without schema in text files, one record per-line. In general, the connector may accept any format that provides an implementation of the Format interface.

Pluggable partitioner

The connector comes out of the box with partitioners that support default partitioning based on Kafka partitions, field partitioning, and time-based partitioning in days or hours. You may implement your own partitioners by extending the Partitioner class. Additionally, you can customize time based partitioning by extending the TimeBasedPartitioner class.

Tip

By default, connectors inherit the partitioner used for the Kafka topic. You can create a custom partitioner for a connector which you must place in the connector’s /lib folder.

You can also put partitioners in a common location of choice. If you choose this option, you must add a symlink to the location from each connector’s /lib folder. For example, you would place a custom partitioner in the path share/confluent-hub-components/partitioners and then add the symlink share/confluent-hub-components/kafka-connect-s3/lib/partitioners -> ../../partitioners.

Non-AWS object storage support

AWS S3 is an industry-standard object storage service. You can use the Kafka Connect S3 connector to connect object storage on non-AWS cloud platforms. For more information, see Using Non-AWS Storage Providers.

• Schema Evolution: Schema evolution only works if the records are generated with the default naming strategy, which is TopicNameStrategy. An error may occur if other naming strategies are used. This is because records are not compatible with each other. schema.compatibility should be set to NONE if other naming strategies are used. This may result in small object files because the sink connector creates a new file every time the schema ID changes between records. See Subject Name Strategy for more information about naming strategies.

Caution

You can’t mix schema and schemaless records in storage using kafka-connect-storage-common. Attempting this causes a runtime exception. If you are using the self-managed version of this connector, this issue will be evident when you review the log files (only available for the self-managed connector).

Writing record keys and headers

The connector is able to write Kafka record keys and headers as files along with the values. Writing record keys and/or headers is optional and may be enabled independently. To enable writing keys set store.kafka.keys=true and to enable headers set store.kafka.headers=true. When writing of the keys or headers is turned on, the connector writes these as additional files. The files have the same name as the file that stores the record values, but end with an extra extension for the respective part of the record, for example .keys.avro and .headers.avro. The files will have a 1-1 mapping with the value files for the records written.

DLQ support: When writing keys or headers is enabled in the connector and the Kafka record has no key or headers present, the record will be written to the DLQ and skipped from the output files.

Note

Schema evolution only works for the record values when this feature is enabled. If the record headers and keys have schemas, and records are sent with a different schema from the initial one, then the connector will currently stop with a failure.

Limitations

• The connector does not currently support Single Message Transformations (SMTs) that modify the topic name. Additionally, the following transformations are not allowed:
  • io.debezium.transforms.ByLogicalTableRouter
  • io.debezium.transforms.outbox.EventRouter
  • org.apache.kafka.connect.transforms.RegexRouter
  • org.apache.kafka.connect.transforms.TimestampRouter
  • io.confluent.connect.transforms.MessageTimestampRouter
  • io.confluent.connect.transforms.ExtractTopic$Key
  • io.confluent.connect.transforms.ExtractTopic$Value
• The connector does not currently support buckets or objects with object locks enabled.

IAM Policy for S3

The AWS user account accessing the S3 bucket must have the following permissions:

• PutObject
• GetObject
• AbortMultipartUpload

Copy the following JSON to create the IAM policy for the user account. Change <bucket-name> to a real bucket name. For more information, see Create and attach a policy to an IAM user. Note that this is the IAM policy for the user account and not a bucket policy.

{
   "Version":"2012-10-17",
   "Statement":[
     {
         "Effect":"Allow",
         "Action":[
           "s3:ListAllMyBuckets"
         ],
         "Resource":"arn:aws:s3:::*"
     },
     {
         "Effect":"Allow",
         "Action":[
           "s3:ListBucket",
           "s3:GetBucketLocation"
         ],
         "Resource":"arn:aws:s3:::<bucket-name>"
     },
     {
         "Effect":"Allow",
         "Action":[
           "s3:PutObject",
           "s3:GetObject",
           "s3:AbortMultipartUpload",
           "s3:PutObjectTagging"
         ],
         "Resource":"arn:aws:s3:::<bucket-name>/*"
     }
   ]
}

Install the Amazon S3 Sink Connector

You can install this connector by using the confluent connect plugin install command, or by manually downloading the ZIP file.

Prerequisites

• You must install the connector on every machine where Connect will run.

• An installation of the latest (latest) connector version.

  To install the latest connector version, navigate to your Confluent Platform installation directory and run the following command:

  confluent connect plugin install confluentinc/kafka-connect-s3:latest
  

  You can install a specific version by replacing latest with a version number as shown in the following example:

  confluent connect plugin install confluentinc/kafka-connect-s3:10.5.0
  

Install the connector manually

Download and extract the ZIP file for your connector and then follow the manual connector installation instructions.

Streaming ETL Demo

To evaluate the Kafka Connect Kinesis Source Connector, AWS S3 Sink Connector, Azure Blob Sink Connector, and Google Cloud GCS Sink Connector in an end-to-end streaming deployment, refer to the Cloud ETL demo. This demo also allows you to evaluate the real-time data processing capabilities of ksqlDB.

License

This connector is available under the Confluent Community License.

Configuration Properties

For a complete list of configuration properties for this connector, see Amazon S3 Sink Connector Configuration Properties.

For an example of how to get Kafka Connect connected to Confluent Cloud, see Connect Self-Managed Kafka Connect to Confluent Cloud.

Mapping Records to S3 Objects

The Amazon S3 Sink connector consumes records from the specified topics, organizes them into different partitions, writes batches of records in each partition to a file, and then uploads those files to the S3 bucket. It uses S3 object paths that include the Kafka topic and partition, the computed partition, and the filename. The S3 connector offers several ways to customize this behavior, including:

• Controlling the names of the S3 objects
• Determining how records are partitioned into S3 objects
• The format used to serialize sets of records into S3 objects
• When to upload S3 objects

S3 Object Names

The S3 data model is a flat structure: each bucket stores objects, and the name of each S3 object serves as the unique key. However, a logical hierarchy can be inferred when the S3 object names uses directory delimiters, such as /. The S3 connector allows you to customize the names of the S3 objects it uploads to the S3 bucket.

In general, the names of the S3 object uploaded by the S3 connector follow this format:

<prefix>/<topic>/<encodedPartition>/<topic>+<kafkaPartition>+<startOffset>.<format>

where:

• <prefix> is specified with the connector’s topics.dir configuration property, which defaults to the literal value topics and helps create uniquely name S3 objects that don’t clash with existing S3 objects in the same bucket.
• <topic> corresponds to the name of the Kafka topic from which the records in this S3 object were read.
• <encodedPartition> is generated by the S3 connector’s partitioner (see Partitioning Records into S3 Objects).
• <kafkaPartition> is the Kafka partition number from which the records in this S3 object were read.
• <startOffset> is the Kafka offset of the first record written to this S3 object.
• <format> is the extension identifing the format in which the records are serialized in this S3 object.

If desired, the / and + characters can be changed using the connector’s directory.delim and file.delim configuration properties.

Partitioning Records into S3 Objects

The S3 connector’s partitioner determines how records read from a Kafka topic are partitioned into S3 objects. The partitioner determines the <encodedPartition> portion of the S3 object names (see S3 Object Names).

The partitioner is specified in the connector configuration with the partitioner.class configuration property. The S3 connector comes with the following partitioners:

• Default Kafka Partitioner: The io.confluent.connect.storage.partitioner.DefaultPartitioner preserves the same topic partitions as in Kafka, and records from each topic partition ultimately end up in S3 objects with names that include the Kafka topic and Kafka partitions. The <encodedPartition> is always <topicName>/partition=<kafkaPartition>, resulting in S3 object names of the form <prefix>/<topic>/partition=<kafkaPartition>/<topic>+<kafkaPartition>+<startOffset>.<format>.
• Field Partitioner: The io.confluent.connect.storage.partitioner.FieldPartitioner determines the partition from the field within each each record identified by the connector’s partition.field.name configuration property, which has no default. This partitioner requires STRUCT record type values. The <encodedPartition> is always <topicName>/<fieldName>=<fieldValue>, resulting in S3 object names of the form <prefix>/<topic>/<fieldName>=<fieldValue>/<topic>+<kafkaPartition>+<startOffset>.<format>.
• Time Based Partitioner: The io.confluent.connect.storage.partitioner.TimeBasedPartitioner determines the partition from the year, month, day, hour, minutes, and/or seconds. This partitioner requires the following connector configuration properties:
  • The path.format configuration property specifies the pattern used for the <encodedPartition> portion of the S3 object name. For example, when path.format='year'=YYYY/'month'=MM/'day'=dd/'hour'=HH, S3 object names will have the form <prefix>/<topic>/year=YYYY/month=MM/day=dd/hour=HH/<topic>+<kafkaPartition>+<startOffset>.<format>.
  • The partition.duration.ms configuration property defines the maximum granularity of the S3 objects within a single encoded partition directory. For example, setting partition.duration.ms=600000 (10 minutes) will result in each S3 object in that directory having no more than 10 minutes of records.
  • The locale configuration property specifies the JDK’s locale used for formatting dates and times. For example, use en-US for US English, en-GB for UK English, fr-FR for French (in France). These may vary by Java version; see the available locales.
  • The timezone configuration property specifies the current timezone in which the dates and times will be treated. Use standard short names for timezones such as UTC or (without daylight savings) PST, EST, and ECT, or longer standard names such as America/Los_Angeles, America/New_York, and Europe/Paris. These may vary by Java version; see the available timezones within each locale, such as those within the “en_US” locale.
  • The timestamp.extractor configuration property determines how to obtain a timestamp from each record. Values can include Wallclock (the default) to use the system time when the record is processed, Record to use the timestamp of the Kafka record denoting when it was produced or stored by the broker, RecordField to extract the timestamp from one of the fields in the record’s value as specified by the timestamp.field configuration property.
• Daily Partitioner: The io.confluent.connect.storage.partitioner.DailyPartitioner is equivalent to the TimeBasedPartitioner with path.format='year'=YYYY/'month'=MM/'day'=dd and partition.duration.ms=86400000 (one day, for one S3 object in each daily directory). This partitioner always results in S3 object names of the form <prefix>/<topic>/year=YYYY/month=MM/day=dd/<topic>+<kafkaPartition>+<startOffset>.<format>. This partitioner requires the following connector configuration properties:
  • The locale configuration property specifies the JDK’s locale used for formatting dates and times. For example, use en-US for US English, en-GB for UK English, fr-FR for French (in France). These may vary by Java version; see the available locales.
  • The timezone configuration property specifies the current timezone in which the dates and times will be treated. Use standard short names for timezones such as UTC or (without daylight savings) PST, EST, and ECT, or longer standard names such as America/Los_Angeles, America/New_York, and Europe/Paris. These may vary by Java version; see the available timezones within each locale, such as those within the “en_US” locale.
  • The timestamp.extractor configuration property determines how to obtain a timestamp from each record. Values can include Wallclock (the default) to use the system time when the record is processed, Record to use the timestamp of the Kafka record denoting when it was produced or stored by the broker, RecordField to extract the timestamp from one of the fields in the record’s value as specified by the timestamp.field configuration property.
• Hourly Partitioner: The io.confluent.connect.storage.partitioner.HourlyPartitioner is equivalent to the TimeBasedPartitioner with path.format='year'=YYYY/'month'=MM/'day'=dd/'hour'=HH and partition.duration.ms=3600000 (one hour, for one S3 object in each hourly directory). This partitioner always results in S3 object names of the form <prefix>/<topic>/year=YYYY/month=MM/day=dd/hour=HH/<topic>+<kafkaPartition>+<startOffset>.<format>. This partitioner requires the following connector configuration properties:
  • The locale configuration property specifies the JDK’s locale used for formatting dates and times. For example, use en-US for US English, en-GB for UK English, fr-FR for French (in France). These may vary by Java version; see the available locales.
  • The timezone configuration property specifies the current timezone in which the dates and times will be treated. Use standard short names for timezones such as UTC or (without daylight savings) PST, EST, and ECT, or longer standard names such as America/Los_Angeles, America/New_York, and Europe/Paris. These may vary by Java version; see the available timezones within each locale, such as those within the “en_US” locale.
  • The timestamp.extractor configuration property determines how to obtain a timestamp from each record. Values can include Wallclock (the default) to use the system time when the record is processed, Record to use the timestamp of the Kafka record denoting when it was produced or stored by the broker, RecordField to extract the timestamp from one of the fields in the record’s value as specified by the timestamp.field configuration property.

As noted below, the choice of timestamp.extractor affects whether the S3 connector can support exactly once delivery.

You can also choose to use a custom partitioner by implementing the io.confluent.connect.storage.partitioner.Partitioner interface, packaging your implementation into a JAR file, and then:

1. Place the JAR inside the libs/ directory in the plugin installation path on each worker node. For example, /usr/share/confluent-hub-components/confluentinc-kafka-s3-10.0.2/libs/.
2. Restart all of the Connect worker nodes.
3. Configure S3 connectors to use your fully-qualified partitioner class name.

S3 Object Formats

The S3 connector can serialize multiple records into each S3 object using a number of formats. The connector’s format.class configuration property identifies the name of the Java class that implements the io.confluent.connect.storage.format.Format interface. The S3 connector comes with several implementations:

• Avro: Use format.class=io.confluent.connect.s3.format.avro.AvroFormat to write the S3 object as an Avro container file that will include the Avro schema in the container file followed by one or more records. The connector’s avro.codec configuration property specifies the Avro compression code, and values can be null (the default) for no Avro compression, deflate to use the deflate algorithm as specified in RFC 1951, snappy to use Google’s Snappy compression library, and bzip2 for BZip2 compression. You can also set enhanced.avro.schema.support=true to enable enum symbol preservation and package name awareness.

• Parquet: Use format.class=io.confluent.connect.s3.format.parquet.ParquetFormat to write the S3 object as a Parquet container file that will include the Parquet schema in the container file. The connector’s parquet.codec configuration property specifies the Parquet compression code, and values can be none (the default) for no Parquet compression, gzip for GZip compression, snappy to use Google’s Snappy compression library, lz4 to use the LZ4 compression algorithm, brotli to use the Brotli algorithm as specified in RFC 7932, zstd to use Facebook’s Zstandard compression algorithm, and lzo to use LZO compression library. You can also set enhanced.avro.schema.support=true to enable enum symbol preservation and package name awareness.

  Important

  • You must use the AvroConverter, ProtobufConverter, or JsonSchemaConverter with ParquetFormat for this connector. Attempting to use the JsonConverter (with or without schemas) results in a NullPointerException and a StackOverflowException.
  • The Amazon S3 Sink connector does not allow recursive schema types. Writing to Parquet output format with a recursive schema type results into a StackOverflowError.

• JSON: Use format.class=io.confluent.connect.s3.format.json.JsonFormat to write the S3 object as a file containing one JSON serialized record per line. The connector’s s3.compression.type configuration property can be set to none (the default) for no compression or gzip for GZip compression.

  Note

  If the source Kafka topic is stored as plain JSON, you cannot use a formatter that requires a schema, you can only use the JSON formatter.

• Raw Bytes: Use format.class=io.confluent.connect.s3.format.bytearray.ByteArrayFormat to write the raw serialized record values delimited with the JDK’s line separator to the S3 object. This requires also using the value.converter=org.apache.kafka.connect.converters.ByteArrayConverter with the connector. Use a different delimiter by specifying the connect’s format.bytearray.separator configuration property.

Caution

There is a risk with using the S3 connector and the ByteArrayFormat and ByteArrayConverter for data containing delimiter characters internally (newlines by default). The risk is that these records may be read back (sourced) incorrectly. For example: You have data that starts out in Avro format. You use the ByteArrayFormat and ByteArrayConverter to sink the data (containing delimiter characters) to S3 storage using the S3 sink connector. When you later source the data to a Kafka topic using the S3 source connector, the original Avro data may be corrupted and potentially unretrievable.

You can also choose to use a custom formatter by implementing the io.confluent.connect.storage.format.Format interface. To do this, package your implementation into a JAR file and then complete the following steps:

1. Place the JAR file into the share/java/kafka-connect-s3 directory of your Confluent Platform installation on each worker node.
2. Restart all of the Connect worker nodes.
3. Configure S3 connectors with format.class set to the fully-qualified class name of your format implementation.

S3 Object Uploads

As the S3 connector processes each record, it uses the partitioner to determine into which encoded partition that record should be written. This continues for each partition until the connector determines that a partition has enough records and should be uploaded to the S3 bucket using the S3 object name for that partition. This technique of knowing when to flush a partition file and upload it to S3 is called the rotation strategy, and there are a number of ways to control this behavior:

• Maximum number of records: The connector’s flush.size configuration property specifies the maximum number of records that should be written to a single S3 object. There is no default for this setting.

• Maximum span of record time: The connector’s rotate.interval.ms specifies the maximum timespan in milliseconds a file can remain open and ready for additional records. The timestamp for each file starts with the record timestamp of the first record written to the file, as determined by the partitioner’s timestamp.extractor. As long as the next record’s timestamp fits within the timespan specified by the rotate.interval.ms, the record will be written to the file. If a record’s timestamp does not fit within the timespan of the file, the connector will flush the file, uploaded it to S3, commit the offsets of the records in that file, and then create a new file with a timespan that starts with the first record and writes the first record to the file.

• Scheduled rotation: The connector’s rotate.schedule.interval.ms specifies the maximum timespan in milliseconds a file can remain open and ready for additional records. Unlike with rotate.interval.ms, with scheduled rotation the timestamp for each file starts with the system time that the first record is written to the file. As long as a record is processed within the timespan specified by rotate.schedule.interval.ms, the record will be written to the file. As soon as a record is processed after the timespan for the current file, the file is flushed, uploaded to S3, and the offset of the records in the file are committed. A new file is created with a timespan that starts with the current system time, and the record is written to the file. The commit will be performed at the scheduled time, regardless of the previous commit time or number of messages. This configuration is useful when you have to commit your data based on current server time, for example at the beginning of every hour. The default value -1 means that this feature is disabled.

  Important

  Be sure to set the timezone configuration property before setting rotate.schedule.interval.ms, otherwise the connector will throw an exception.

These strategies can be combined as needed, and rotation occurs whenever any of the strategies signals a rotation.

The first strategy will cause a rotation as soon as enough records have been written to the file, and can be calculated after each record has been written to the file. In other words, the file can be closed and uploaded to S3 as soon as it is full.

When using rotate.interval.ms, the connector only closes and uploads a file to S3 when the next file does not belong based upon that record’s timestamp. In other words, if the connector has no more records to process, the connector may keep the file open until the connector can process another record (this can be a long time).

Scheduled rotation uses rotate.schedule.interval.ms to close the file and upload to S3 on a regular basis using the current time, rather than the record time. Even if the connector has no more records to process, Connect will still call the connector at least every offset.flush.interval.ms as defined in the Connect worker’s configuration file. And every time this occurs, the connector uses the current time to determine if the currently opened file should be closed and uploaded to S3.

Note

Not all rotation strategies are compatible with the S3 connector’s ability to deliver S3 objects exactly once with eventual consistency. See the Exactly Once section below for details.

The S3 object uploaded by the connector can be quite large, and the connector supports using a multi-part upload mechanism. The s3.part.size configuration property defaults to 26214400 bytes (25MB), and specifies the maximum size of each S3 object part used to upload a single S3 object.

Additionally, the schema.compatibility setting (see Schema Evolution) will also affect when one file is closed and uploaded to an S3 object. If a record cannot be written to one file because its schema has changed relative to the records already in the file, the connector will rotate by closing the file, uploading it to S3, committing offsets for the records in the file, creating a new file and writing the new record.

Exactly-once delivery on top of eventual consistency

The S3 connector is able to provide exactly-once semantics to consumers of the objects it exports to S3, under the condition that the connector is supplied with a deterministic partitioner.

Currently, out of the available partitioners, the default and field partitioners are always deterministic. TimeBasedPartitioner can be deterministic with some configurations, discussed below. This implies that, when any of these partitioners is used, splitting of files always happens at the same offsets for a given set of Kafka records. These partitioners take into account flush.size and schema.compatibility to decide when to roll and save a new file to S3. The connector always delivers files in S3 that contain the same records, even under the presence of failures. If a connector task fails before an upload completes, the file does not become visible to S3. If, on the other hand, a failure occurs after the upload has completed but before the corresponding offset is committed to Kafka by the connector, then a re-upload will take place. However, such a re-upload is transparent to the user of the S3 bucket, who at any time will have access to the same records made eventually available by successful uploads to S3.

To guarantee exactly-once semantics with the TimeBasedPartitioner, the connector must be configured to use a deterministic implementation of TimestampExtractor and a deterministic rotation strategy. The deterministic timestamp extractors are Kafka records (timestamp.extractor=Record) or record fields (timestamp.extractor=RecordField). The deterministic rotation strategy configuration is rotate.interval.ms (setting rotate.schedule.interval.ms is nondeterministic and will invalidate exactly-once guarantees).

Schema Evolution

Important

Schema evolution only works if the records are generated with the default naming strategy, which is TopicNameStrategy. An error may occur if other naming strategies are used. This is because records are not compatible with each other. schema.compatibility should be set to NONE if other naming strategies are used. This may result in small object files because the sink connector creates a new file every time the schema ID changes between records. See Subject Name Strategy for more information about naming strategies.

The S3 connector supports schema evolution and reacts to schema changes of data according to the schema.compatibility configuration. In this section, we will explain how the connector reacts to schema evolution under different values of schema.compatibility. The schema.compatibility can be set to NONE, BACKWARD, FORWARD and FULL, which means NO compatibility, BACKWARD compatibility, FORWARD compatibility and FULL compatibility respectively.

• NO Compatibility: By default, the schema.compatibility is set to NONE. In this case, the connector ensures that each file written to S3 has the proper schema. When the connector observes a schema change in data, it commits the current set of files for the affected topic partitions and writes the data with new schema in new files.

• BACKWARD Compatibility: If a schema is evolved in a backward compatible way, we can always use the latest schema to query all the data uniformly. For example, removing fields is backward compatible change to a schema, since when we encounter records written with the old schema that contain these fields we can just ignore them. Adding a field with a default value is also backward compatible.

  If BACKWARD is specified in the schema.compatibility, the connector keeps track of the latest schema used in writing data to S3, and if a data record with a schema version larger than current latest schema arrives, the connector commits the current set of files and writes the data record with new schema to new files. For data records arriving at a later time with schema of an earlier version, the connector projects the data record to the latest schema before writing to the same set of files in S3.

• FORWARD Compatibility: If a schema is evolved in a forward compatible way, we can always use the oldest schema to query all the data uniformly. Removing a field that had a default value is forward compatible, since the old schema will use the default value when the field is missing.

  If FORWARD is specified in the schema.compatibility, the connector projects the data to the oldest schema before writing to the same set of files in S3.

• FULL Compatibility: Full compatibility means that old data can be read with the new schema and new data can also be read with the old schema.

  If FULL is specified in the schema.compatibility, the connector performs the same action as BACKWARD.

Schema evolution in the S3 connector works in the same way as in the HDFS connector.

Automatic Retries

The S3 connector may experience problems writing to the S3 bucket, due to network partitions, interruptions, or even AWS throttling limits. In many cases, the connector will retry the request a number of times before failing. To prevent from further overloading the network or S3 service, the connector uses an exponential backoff technique to give the network and/or service time to recover. The technique adds randomness, called jitter, to the calculated backoff times to prevent a thundering herd, where large numbers of requests from many tasks are submitted concurrently and overwhelm the service. Randomness spreads out the retries from many tasks and should reduce the overall time required to complete all outstanding requests compared to simple exponential backoff. The goal is to spread out the requests to S3 as much as possible.

The maximum number of retry attempts is dictated by the s3.part.retries S3 connector configuration property, which defaults to three attempts. The delay for retries is dependent upon the connector’s s3.retry.backoff.ms configuration property, which defaults to 200 milliseconds. The actual delay is randomized, but the maximum delay can be calculated as a function of the number of retry attempts with ${s3.retry.backoff.ms} * 2 ^ (retry-1), where retry is the number of attempts taken so far in the current iteration. In order to keep the maximum delay within a reasonable duration, it is capped at 24 hours. For example, the following table shows the possible wait times before submitting each of the three retry attempts.

Range of backoff times for each retry using the default configuration

Retry	Minimum Backoff (sec)	Maximum Backoff (sec)	Total Potential Delay from First Attempt (sec)
1	0.0	0.2	0.2
2	0.0	0.4	0.6
3	0.0	0.8	1.4

Increasing the maximum number of retries adds more backoff:

Range of backoff times for additional retries

Retry	Minimum Backoff (sec)	Maximum Backoff (sec)	Total Potential Delay from First Attempt (sec)
4	0.0	1.6	3.0
5	0.0	3.2	6.2
6	0.0	6.4	12.6
7	0.0	12.8	25.4
8	0.0	25.6	51.0
9	0.0	51.2	102.2
10	0.0	102.4	204.6

At some point, maximum backoff time will reach saturation and will be capped at 24 hours. From the example below, all attempts starting with 20 will have maximum backoff time as 24 hours:

Range of backoff times when reaching the cap of 24 hours

Retry	Minimum Backoff (sec)	Maximum Backoff (sec)	Total Potential Delay from First Attempt (sec)
15	0.0	3276.8	6553.4
16	0.0	6553.6	13107.0
17	0.0	13107.2	26214.2
18	0.0	26214.4	52428.6
19	0.0	52428.8	104857.4
20	0.0	86400.0	191257.4
21	0.0	86400.0	277657.4

It’s not advised to set s3.part.retries too high since making more attempts after reaching a cap of 24 hours isn’t practical. You can adjust both the s3.part.retries and s3.retry.backoff.ms connector configuration properties to achieve the desired retry and backoff characteristics.

AWS Credentials

The following sections provide information about how to configure an S3 connector to provide credentials when connecting to AWS.

Credentials provider chain

By default, the S3 connector looks for S3 credentials in the following locations and in the following order:

1. The AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY environment variables accessible to the Connect worker processes where the connector will be deployed. These variables are recognized by the AWS CLI and all AWS SDKs (except for the AWS SDK for .NET). You use export to set these variables.

   export AWS_ACCESS_KEY_ID=<your_access_key_id>
   export AWS_SECRET_ACCESS_KEY=<your_secret_access_key>
   

   The AWS_ACCESS_KEY and AWS_SECRET_KEY can be used instead, but are not recognized by the AWS CLI.

2. The aws.accessKeyId and aws.secretKey Java system properties on the Connect worker processes where the connector will be deployed. However, these variables are only recognized by the AWS SDK for Java and are not recommended.

3. The ~/.aws/credentials file located in the home directory of the operating system user that runs the Connect worker processes. These credentials are recognized by most AWS SDKs and the AWS CLI. Use the following AWS CLI command to create the credentials file:

   aws configure
   

   You can also manually create the credentials file using a text editor. The file should contain lines in the format shown in the example below. See AWS Credentials File Format for additional details.

   [default]
   aws_access_key_id = <your_access_key_id>
   aws_secret_access_key = <your_secret_access_key>
   

   Note

   When creating the credentials file, make sure that the user creating the credentials file is the same user that runs the Connect worker processes and that the credentials file is in this user’s home directory. Otherwise, the S3 connector will not be able to find the credentials.

4. A query sent to http://169.254.170.2${AWS_CONTAINER_CREDENTIALS_RELATIVE_URI} to return AWS credentials. This is applicable only if the Connect worker processes are running in AWS containers.

5. A metadata query that returns credentials from an EC2 instance. This is applicable only if the Connect worker processes are running in EC2 instances.

Choose one of the above to define the AWS credentials that the S3 connectors use, verify the credentials implementation is set correctly, and then restart all of the Connect worker processes.

Note

Confluent recommends using either Environment variables or a Credentials file because these are the most straightforward, and they can be checked using the AWS CLI tool before running the connector.

All S3 connectors run in a single Connect worker cluster and use the same credentials. This is sufficient for many use cases. If you want more control, refer to the following section to learn more about controlling and customizing how the S3 connector gets AWS credentials.

Caution

If you configure one of the AWS key and AWS secret key implementations (as detailed above), credentials can not be supplied through the following credentials providers or by using the Trusted Account Credentials implementation. Attempting to provide credentials using multiple implementations will cause authentication failure.

Credentials providers

A credentials provider is a Java class that implements the com.amazon.auth.AWSCredentialsProvider interface in the AWS Java library and returns AWS credentials from the environment. By default the S3 connector configuration property s3.credentials.provider.class uses the com.amazon.auth.DefaultAWSCredentialsProviderChain class. This class and interface implementation chains together five other credential provider classes. The DefaultAWSCredentialsProviderChain implementation looks for credentials in the following order:

1. Environment variables using the EnvironmentVariableCredentialsProvider class implementation. This implementation uses environment variables AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY. Environment variables AWS_ACCESS_KEY and AWS_SECRET_KEY are also supported by this implementation; however, these two variables are only recognized by the AWS SDK for Java and are not recommended.

2. Java system properties using the SystemPropertiesCredentialsProvider class implementation. This implementation uses Java system properties aws.accessKeyId and aws.secretKey.

3. Credentials file using the ProfileCredentialsProvider class implementation. This implementation uses a credentials file located in the path ~/.aws/credentials. This credentials provider can be used by most AWS SDKs and the AWS CLI. Use the following AWS CLI command to create the credentials file:

   aws configure
   

   You can also manually create the credentials file using a text editor. The file should contain lines in the format shown in the example below. See AWS Credentials File Format for additional details.

   [default]
   aws_access_key_id = <your_access_key_id>
   aws_secret_access_key = <your_secret_access_key>
   

   Note

   When creating the credentials file, make sure that the user creating the credentials file is the same user that runs the Connect worker processes and that the credentials file is in this user’s home directory. Otherwise, the S3 connector will not be able to find the credentials.

4. Amazon Elastic Container Service (ECS) container credentials using the ContainerCredentialsProvider class implementation. This implementation uses a query sent to http://169.254.170.2${AWS_CONTAINER_CREDENTIALS_RELATIVE_URI} to return AWS credentials for the S3 connector. For this provider to work, the environment variable AWS_CONTAINER_CREDENTIALS_RELATIVE_URI must be set. See IAM Roles for Tasks for additional information about setting up this query.

5. EC2 instance profile credentials using the InstanceProfileCredentialsProvider class implementation. EC2 instance metadata is queried for credentials. See Amazon EC2 metadata service for additional information about instance metadata queries. See Working with AWS credentials for additional information and updates from AWS.

   Note

   EC2 instance profile credentials can be used only if the environment variable AWS_CONTAINER_CREDENTIALS_RELATIVE_URI is not set. See EC2ContainerCredentialsProviderWrapper for more information.

Using trusted account credentials

This connector can assume a role and use credentials from a separate trusted account. This is a default feature provided with recent versions of this connector that include an updated version of the AWS SDK.

Important

You cannot use assumed role credentials to access AWS through a proxy server without first passing environment variables or system properties. This is due to an AWS SDK limitation.

After you create the trust relationship, an IAM user or an application from the trusted account can use the AWS Security Token Service (AWS STS) AssumeRole API operation. This operation provides temporary security credentials that enable access to AWS resources for the connector. For details, see Creating a Role to Delegate Permissions to an IAM User.

Example:

Profile in ~/.aws/credentials:

[default]
role_arn=arn:aws:iam::037803949979:role/kinesis_cross_account_role
source_profile=staging
role_session_name = OPTIONAL_SESSION_NAME

[staging]
aws_access_key_id = <STAGING KEY>
aws_secret_access_key = <STAGING SECRET>

To allow the connector to assume a role with the right permissions, set the Amazon Resource Name (ARN) for this role. Additionally, you must choose between source_profile or credential_source as the way to get credentials that have permission to assume the role, in the environment where the connector is running.

Note

When setting up trusted account credentials, be aware that the approach of loading profiles from both ~/.aws/credentials and ~/.aws/config does not work when configuring this connector. Assumed role settings and credentials must be placed in the ~/.aws/credentials file.

Additionally, the S3 sink connector implements the AwsAssumeRoleCredentialsProvider which means you can use the following configs to configure the assume role operation.

s3.credentials.provider.class=io.confluent.connect.s3.auth.AwsAssumeRoleCredentialsProvider
s3.credentials.provider.sts.role.arn=arn:aws:iam::012345678901:role/my-restricted-role
s3.credentials.provider.sts.role.session.name=session-name
s3.credentials.provider.sts.role.external.id=external-id

Using other implementations

You can use a different credentials provider. To do this, set the s3.credentials.provider.class property to the name of any class that implements the com.amazon.auth.AWSCredentialsProvider interface.

Complete the following steps to use a different credentials provider:

1. Find or create a Java credentials provider class that implements the com.amazon.auth.AWSCredentialsProvider interface.

2. Put the class file in a JAR file.

3. Place the JAR file in the share/java/kafka-connect-s3 directory on all Connect workers.

4. Restart the Connect workers.

5. Change the S3 connector property file to use your custom credentials. Add the provider class entry s3.credentials.provider.class=<className> in the S3 connector properties file.

   Important

   You must use the fully qualified class name in the <className> entry.

Quick Start

In this quick start, we use the S3 connector to export data produced by the Avro console producer to S3.

Before you begin, create an AWS S3 destination bucket and grant write access to the user or IAM role completing these procedures. See Identity and access management in Amazon S3 for more information.

Next, start the services with one command using the Confluent CLI confluent local commands:

Tip

If not already in your PATH, add Confluent’s bin directory by running: export PATH=<path-to-confluent>/bin:$PATH

confluent local services start

Every service will start in order, printing a message with its status:

Starting Zookeeper
Zookeeper is [UP]
Starting Kafka
Kafka is [UP]
Starting Schema Registry
Schema Registry is [UP]
Starting Kafka REST
Kafka REST is [UP]
Starting Connect
Connect is [UP]
Starting KSQL Server
KSQL Server is [UP]
Starting Control Center
Control Center is [UP]

Note

Make sure the S3 connector has write access to the S3 bucket shown in s3.bucket.name and can deploy credentials successfully. See AWS Credentials for detailed information about setting up credential providers.

To import a few records with a simple schema in Kafka, start the Avro console producer as follows:

./bin/kafka-avro-console-producer --broker-list localhost:9092 --topic s3_topic \
--property value.schema='{"type":"record","name":"myrecord","fields":[{"name":"f1","type":"string"}]}'

Then, in the console producer, type in:

{"f1": "value1"}
{"f1": "value2"}
{"f1": "value3"}
{"f1": "value4"}
{"f1": "value5"}
{"f1": "value6"}
{"f1": "value7"}
{"f1": "value8"}
{"f1": "value9"}

The nine records entered are published to the Kafka topic s3_topic in Avro format.

Before starting the connector, make sure that the configurations in etc/kafka-connect-s3/quickstart-s3.properties are properly set to your configurations of S3, for example s3.bucket.name points to your bucket, s3.region directs to your S3 region and flush.size=3 for this example. Then start the S3 connector by loading its configuration with the following command:

  confluent local services connect connector load s3-sink

{
  "name": "s3-sink",
  "config": {
    "connector.class": "io.confluent.connect.s3.S3SinkConnector",
    "tasks.max": "1",
    "topics": "s3_topic",
    "s3.region": "us-west-2",
    "s3.bucket.name": "confluent-kafka-connect-s3-testing",
    "s3.part.size": "5242880",
    "flush.size": "3",
    "storage.class": "io.confluent.connect.s3.storage.S3Storage",
    "format.class": "io.confluent.connect.s3.format.avro.AvroFormat",
    "schema.generator.class": "io.confluent.connect.storage.hive.schema.DefaultSchemaGenerator",
    "partitioner.class": "io.confluent.connect.storage.partitioner.DefaultPartitioner",
    "schema.compatibility": "NONE",
    "name": "s3-sink"
  },
  "tasks": []
}

To check that the connector started successfully, view the Connect worker’s log by running:

confluent local services connect log

Towards the end of the log you should see that the connector starts, logs a few messages, and then uploads data from Kafka to S3. Once the connector has ingested some records check that the data is available in S3, for instance by using AWS CLI:

aws s3api list-objects --bucket "your-bucket-name"

You should see three objects with keys:

topics/s3_topic/partition=0/s3_topic+0+0000000000.avro
topics/s3_topic/partition=0/s3_topic+0+0000000003.avro
topics/s3_topic/partition=0/s3_topic+0+0000000006.avro

Note

The S3 connector doesn’t use the message key. If you need to store the key in the S3 objects and information in the key doesn’t already exist in the value, use a custom transformation with the connector to add the message key to the value.

Each file is encoded as <topic>+<kafkaPartition>+<startOffset>.<format>.

To verify the contents, copy each file from S3 to your local filesystem by running the following command:

aws s3 cp s3://<your-bucket>/topics/s3_topic/partition=0/s3_topic+0+0000000000.avro

To print the records, use avro-tools-1.8.2.jar (available in the Apache Archives):

java -jar avro-tools-1.8.2.jar tojson s3_topic+0+0000000000.avro

For the previous file, you should see the following output (with the rest of the records contained in the other two files):

{"f1":"value1"}
{"f1":"value2"}
{"f1":"value3"}

Finally, stop the Connect worker as well as all the rest of the Confluent services by running:

confluent local stop

Your output should resemble:

Stopping Control Center
Control Center is [DOWN]
Stopping KSQL Server
KSQL Server is [DOWN]
Stopping Connect
Connect is [DOWN]
Stopping Kafka REST
Kafka REST is [DOWN]
Stopping Schema Registry
Schema Registry is [DOWN]
Stopping Kafka
Kafka is [DOWN]
Stopping Zookeeper
Zookeeper is [DOWN]

Or, stop all the services and additionally wipe out any data generated during this quick start by running:

confluent local destroy

Your output should resemble:

Stopping Control Center
Control Center is [DOWN]
Stopping KSQL Server
KSQL Server is [DOWN]
Stopping Connect
Connect is [DOWN]
Stopping Kafka REST
Kafka REST is [DOWN]
Stopping Schema Registry
Schema Registry is [DOWN]
Stopping Kafka
Kafka is [DOWN]
Stopping Zookeeper
Zookeeper is [DOWN]
Deleting: /var/folders/ty/rqbqmjv54rg_v10ykmrgd1_80000gp/T/confluent.PkQpsKfE

Example Property File Settings

Refer to the following examples for information about setting up the connector configuration.

Basic example

The example settings are contained in etc/kafka-connect-s3/quickstart-s3.properties as follows:

name=s3-sink
connector.class=io.confluent.connect.s3.S3SinkConnector
tasks.max=1
topics=s3_topic
flush.size=3

The first few settings are common to most connectors. topics specifies the topics we want to export data from, in this case s3_topic. The property flush.size specifies the number of records per partition the connector needs to write before completing a multipart upload to S3.

s3.bucket.name=confluent-kafka-connect-s3-testing
s3.part.size=5242880

The next settings are specific to AWS S3. A mandatory setting is the name of your S3 bucket to host the exported Kafka records. Other useful settings are s3.region, which you should set if you use a region other than the default, and s3.part.size to control the size of each part in the multipart uploads that will be used to upload a single chunk of Kafka records.

storage.class=io.confluent.connect.s3.storage.S3Storage
format.class=io.confluent.connect.s3.format.avro.AvroFormat
schema.generator.class=io.confluent.connect.storage.hive.schema.DefaultSchemaGenerator
partitioner.class=io.confluent.connect.storage.partitioner.DefaultPartitioner

These class settings are required to specify the storage interface (here S3), the output file format, currently io.confluent.connect.s3.format.avro.AvroFormat or io.confluent.connect.s3.format.json.JsonFormat and the partitioner class along with its schema generator class. When using a format with no schema definition, it is sufficient to set the schema generator class to its default value.

schema.compatibility=NONE

Finally, schema evolution is disabled in this example by setting schema.compatibility to NONE, as explained above.

For detailed descriptions for all the available configuration options of the S3 connector go to Amazon S3 Sink Connector Configuration Properties.

Write raw message values into S3

It is possible to use the S3 connector to write out the unmodified original message values into newline-separated files in S3. We accomplish this by telling Connect to not deserialize any of the messages, and by configuring the S3 connector to store the message values in a binary format in S3.

The first part of our S3 connector is similar to other examples:

name=s3-raw-sink
connector.class=io.confluent.connect.s3.S3SinkConnector
tasks.max=1
topics=s3_topic
flush.size=3

The topics setting specifies the topics we want to export data from, in this case s3_topic. The property flush.size specifies the number of records per partition the connector needs to write before completing a multipart upload to S3.

Next we need to configure the particulars of AWS S3:

s3.bucket.name=confluent-kafka-connect-s3-testing
s3.region=us-west-2
s3.part.size=5242880
s3.compression.type=gzip

The s3.bucket.name is mandatory and names your S3 bucket where the exported Kafka records should be written. Another useful setting is s3.region that you should set if you use a region other than the default. And since the S3 connector uses multi-part uploads, you can use the s3.part.size to control the size of each of these continuous parts used to upload Kafka records into a single S3 object. The part size affects throughput and latency, as an S3 object is visible/available only after all parts are uploaded. The s3.compression.type specifies that we want the S3 connector to compress our S3 objects using GZIP compression, adding the .gz extension to any files (see below).

So far this example configuration is relatively typical of most S3 connectors. Now lets define that we should read the raw message values and write them in binary format:

value.converter=org.apache.kafka.connect.converters.ByteArrayConverter
format.class=io.confluent.connect.s3.format.bytearray.ByteArrayFormat
storage.class=io.confluent.connect.s3.storage.S3Storage
schema.compatibility=NONE

The value.converter setting overrides for our connector the default that is in the Connect worker configuration, and we use the ByteArrayConverter to instruct Connect to skip deserializing the message values and instead give the connector the message values in their raw binary form. We use the format.class setting to instruct the S3 connector to write these binary message values as-is into S3 objects. By default the message values written to the same S3 object will be separated by a newline character sequence, but you can control this with the format.bytearray.separator setting, and you may want to consider this if your messages might contain newlines. Also, by default the files written to S3 will have an extension of .bin (before compression, if enabled), or you can use the format.bytearray.extension setting to change the pre-compression filename extension.

Next we need to decide how we want to partition the consumed messages in S3 objects. We have a few options, including the default partitioner that preserves the same partitions as in Kafka:

partitioner.class=io.confluent.connect.storage.partitioner.DefaultPartitioner

Or, we could instead partition by the timestamp of the Kafka messages:

partitioner.class=io.confluent.connect.storage.partitioner.TimeBasedPartitioner
timestamp.extractor=Record

or the timestamp that the S3 connector processes each message:

partitioner.class=io.confluent.connect.storage.partitioner.TimeBasedPartitioner
timestamp.extractor=Wallclock

Custom partitioners are always an option, too. Just be aware that since the record value is an opaque binary value, we cannot extract timestamps from fields using the RecordField option.

The S3 connector configuration outlined above results in newline-delimited gzipped objects in S3 with .bin.gz.

Using Non-AWS Storage Providers

Many cloud providers implement an AWS S3-compatible API. You can use the Kafka Connect S3 connector to connect to object storage on their platform. When configuring the S3 connector for object storage on other cloud providers, include the following configuration option (if applicable for the cloud provider):

store.url

The object storage connection URL.

• Type: string
• Default: null
• Importance: high

Important

Any AWS S3-compatible API you use must support multi-part uploads for the Kafka Connect S3 connector. See Multipart Upload Overview for more information.

Suggested Reading

Blog post From Apache Kafka to Amazon S3: Exactly Once