kinesis-s3-recovery

Recovers Kinesis Firehose source record backups from S3, replaying them into the stream.

Installation

The python scripts make use of python 3 and boto3. pipenv install or pip install boto3 will do.

Deployment

A package.sh shell script will create a lambdas.zip that can be uploaded to AWS Lambda. The function handlers are:

• queue_firehose_s3_backups.main
• recover_firehose_s3_backup.main

The lambda functions will need an IAM role that has permissions to:

• Read from S3
• Put to Kinesis Firehose
• Send and receive messages from SQS

Functions

Files are recovered in two steps, with two lambda functions:

1. queue_firehose_s3_backups.py lists all the files in the S3 source records backup and SQS queues each file
2. recover_firehose_s3_backup.py is triggered by the SQS queue. It parses objects from each file and puts records to the Kinesis Firehose stream

Test

With the lambda functions deployed, they can be tested with an event like the following

queue_firehose_s3_backups.py

{
  "bucket": "my-s3-bucket",
  "prefix": "is/optional/",
  "queue_url": "https://sqs.us-east-1.amazonaws.com/123456789/my_queue.fifo",
  "kinesis_stream": "my-kinesis-stream"
}

The file is assumed to be formatted in the idiosyncratic Kinesis Firehose manner, with objects concatenated together.

Credits

Thanks to Tom Chapin for the code to parse the Kinesis Firehose S3 record format: https://stackoverflow.com/questions/34468319/reading-the-data-written-to-s3-by-amazon-kinesis-firehose-stream

License

MIT