By default, you can create up to 20 delivery streams per region.
This is an asynchronous operation that immediately returns. The initial status of the delivery stream is CREATING
. After the delivery stream is created, its status is ACTIVE
and it now accepts data. Attempts to send data to a delivery stream that is not in the ACTIVE
state cause an exception. To check the state of a delivery stream, use DescribeDeliveryStream.
A delivery stream is configured with a single destination: Amazon S3, Amazon Elasticsearch Service, or Amazon Redshift. You must specify only one of the following destination configuration parameters: ExtendedS3DestinationConfiguration, S3DestinationConfiguration, ElasticsearchDestinationConfiguration, or RedshiftDestinationConfiguration.
When you specify S3DestinationConfiguration, you can also provide the following optional values: BufferingHints, EncryptionConfiguration, and CompressionFormat. By default, if no BufferingHints value is provided, Firehose buffers data up to 5 MB or for 5 minutes, whichever condition is satisfied first. Note that BufferingHints is a hint, so there are some cases where the service cannot adhere to these conditions strictly; for example, record boundaries are such that the size is a little over or under the configured buffering size. By default, no encryption is performed. We strongly recommend that you enable encryption to ensure secure data storage in Amazon S3.
A few notes about Amazon Redshift as a destination:
An Amazon Redshift destination requires an S3 bucket as intermediate location, as Firehose first delivers data to S3 and then uses COPY
syntax to load data into an Amazon Redshift table. This is specified in the RedshiftDestinationConfiguration.S3Configuration parameter.
The compression formats SNAPPY
or ZIP
cannot be specified in RedshiftDestinationConfiguration.S3Configuration because the Amazon Redshift COPY
operation that reads from the S3 bucket doesn't support these compression formats.
We strongly recommend that you use the user name and password you provide exclusively with Firehose, and that the permissions for the account are restricted for Amazon Redshift INSERT
permissions.
Firehose assumes the IAM role that is configured as part of the destination. The role should allow the Firehose principal to assume the role, and the role should have permissions that allows the service to deliver the data. For more information, see Amazon S3 Bucket Access in the Amazon Kinesis Firehose Developer Guide.