Custom Prefixes for Amazon S3 Objects - Amazon Kinesis Data Firehose
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Custom Prefixes for Amazon S3 Objects

If your destination is Amazon S3, Amazon OpenSearch Service, or Splunk, you can configure the Amazon S3 object keys used for delivering data from Kinesis Data Firehose. To do this, you specify expressions that Kinesis Data Firehose evaluates at delivery time. The final object keys have the format <evaluated prefix><suffix>, where the suffix has the format <delivery stream name>-<delivery stream version>-<year>-<month>-<day>-<hour>-<minute>-<second>-<uuid><file extension>. You can't change the suffix field.

You can use expressions of the following forms in your custom prefix: !{namespace:value}, where namespace can be one of the following, as explained in the following sections.

  • firehose

  • timestamp

  • partitionKeyFromQuery

  • partitionKeyFromLambda

If a prefix ends with a slash, it appears as a folder in the Amazon S3 bucket. For more information, see Amazon S3 Object Name Format in the Amazon Kinesis Data Firehose Developer Guide.

The timestamp namespace

Valid values for this namespace are strings that are valid Java DateTimeFormatter strings. As an example, in the year 2018, the expression !{timestamp:yyyy} evaluates to 2018.

When evaluating timestamps, Kinesis Data Firehose uses the approximate arrival timestamp of the oldest record that's contained in the Amazon S3 object being written.

The timestamp is always in UTC.

If you use the timestamp namespace more than once in the same prefix expression, every instance evaluates to the same instant in time.

The firehose namespace

There are two values that you can use with this namespace: error-output-type and random-string. The following table explains how to use them.

The firehose namespace values
Conversion Description Example input Example output Notes
error-output-type Evaluates to one of the following strings, depending on the configuration of your delivery stream, and the reason of failure: {processing-failed, AmazonOpenSearchService-failed, splunk-failed, format-conversion-failed, http-endpoint-failed}.

If you use it more than once in the same expression, every instance evaluates to the same error string..

myPrefix/result=!{firehose:error-output-type}/!{timestamp:yyyy/MM/dd} myPrefix/result=processing-failed/2018/08/03 The error-output-type value can only be used in the ErrorOutputPrefix field.

Evaluates to a random string of 11 characters. If you use it more than once in the same expression, every instance evaluates to a new random string.

myPrefix/!{firehose:random-string}/ myPrefix/046b6c7f-0b/ You can use it with both prefix types.

You can place it at the beginning of the format string to get a randomized prefix, which is sometimes necessary for attaining extremely high throughput with Amazon S3.

partitionKeyFromLambda and partitionKeyFromQuery namespaces

For dynamic partitioning, you must use the following expression format in your S3 bucket prefix: !{namespace:value}, where namespace can be either partitionKeyFromQuery or partitionKeyFromLambda, or both. If you are using inline parsing to create the partitioning keys for your source data, you must specify an S3 bucket prefix value that consists of expressions specified in the following format: "partitionKeyFromQuery:keyID". If you are using an Amazon Lambda function to create partitioning keys for your source data, you must specify an S3 bucket prefix value that consists of expressions specified in the following format: "partitionKeyFromLambda:keyID". For more information, see the "Choose Amazon S3 for Your Destination" in Creating an Amazon Kinesis Data Firehose Delivery Stream.

Semantic rules

The following rules apply to Prefix and ErrorOutputPrefix expressions.

  • For the timestamp namespace, any character that isn't in single quotes is evaluated. In other words, any string escaped with single quotes in the value field is taken literally.

  • If you specify a prefix that doesn't contain a timestamp namespace expression, Kinesis Data Firehose appends the expression !{timestamp:yyyy/MM/dd/HH/}to the value in the Prefix field.

  • The sequence !{ can only appear in !{namespace:value} expressions.

  • ErrorOutputPrefix can be null only if Prefix contains no expressions. In this case, Prefix evaluates to <specified-prefix>yyyy/MM/DDD/HH/ and ErrorOutputPrefix evaluates to <specified-prefix><error-output-type>YYYY/MM/DDD/HH/. DDD represents the day of the year.

  • If you specify an expression for ErrorOutputPrefix, you must include at least one instance of !{firehose:error-output-type}.

  • Prefix can't contain !{firehose:error-output-type}.

  • Neither Prefix nor ErrorOutputPrefix can be greater than 512 characters after they're evaluated.

  • If the destination is Amazon Redshift, Prefix must not contain expressions and ErrorOutputPrefix must be null.

  • When the destination is Amazon OpenSearch Service or Splunk, and no ErrorOutputPrefix is specified, Kinesis Data Firehose uses the Prefix field for failed records.

  • When the destination is Amazon S3, the Prefix and ErrorOutputPrefix in the Amazon S3 destination configuration are used for successful records and failed records, respectively. If you use the Amazon CLI or the API, you can use ExtendedS3DestinationConfiguration to specify an Amazon S3 backup configuration with its own Prefix and ErrorOutputPrefix.

  • When you use the Amazon Web Services Management Console and set the destination to Amazon S3, Kinesis Data Firehose uses the Prefix and ErrorOutputPrefix in the destination configuration for successful records and failed records, respectively. If you specify a prefix but no error prefix, Kinesis Data Firehose automatically sets the error prefix to !{firehose:error-output-type}/.

  • When you use ExtendedS3DestinationConfiguration with the Amazon CLI, the API, or Amazon CloudFormation, if you specify a S3BackupConfiguration, Kinesis Data Firehose doesn't provide a default ErrorOutputPrefix.

  • You cannot use partitionKeyFromLambda and partitionKeyFromQuery namespaces when creating ErrorOutputPrefix expressions.

Example prefixes

Prefix and ErrorOutputPrefix examples
Input Evaluated prefix (at 10:30 AM UTC on Aug 27, 2018)

Prefix: Unspecified

ErrorOutputPrefix: myFirehoseFailures/!{firehose:error-output-type}/

Prefix: 2018/08/27/10

ErrorOutputPrefix: myFirehoseFailures/processing-failed/

Prefix: !{timestamp:yyyy/MM/dd}

ErrorOutputPrefix: Unspecified

Invalid input: ErrorOutputPrefix can't be null when Prefix contains expressions

Prefix: myFirehose/DeliveredYear=!{timestamp:yyyy}/anyMonth/rand=!{firehose:random-string}

ErrorOutputPrefix: myFirehoseFailures/!{firehose:error-output-type}/!{timestamp:yyyy}/anyMonth/!{timestamp:dd}

Prefix: myFirehose/DeliveredYear=2018/anyMonth/rand=5abf82daaa5

ErrorOutputPrefix: myFirehoseFailures/processing-failed/2018/anyMonth/10

Prefix: myPrefix/year=!{timestamp:yyyy}/month=!{timestamp:MM}/day=!{timestamp:dd}/hour=!{timestamp:HH}/

ErrorOutputPrefix: myErrorPrefix/year=!{timestamp:yyyy}/month=!{timestamp:MM}/day=!{timestamp:dd}/hour=!{timestamp:HH}/!{firehose:error-output-type}

Prefix: myPrefix/year=2018/month=07/day=06/hour=23/

ErrorOutputPrefix: myErrorPrefix/year=2018/month=07/day=06/hour=23/processing-failed

Prefix: myFirehosePrefix

ErrorOutputPrefix: Unspecified

Prefix: myFirehosePrefix/2018/08/27/

ErrorOutputPrefix: myFirehosePrefix/processing-failed/2018/08/27/