bert-etl

Bert

A microframework for simple ETL solutions.

Architecture

At its core, bert-etl uses Dynamodb Streams to communicate between lambda functions. bert-etl.yaml provides control on how the initial lambda function is called, either by periodic events, sns topics, or s3 bucket (planned)events. Passing an event to bert-etl is straight forward from zappa or a generic AWS lambda function you've hooked up to API Gateway.

At this moment in time, there are no plans to attach API Gateway to bert-etl.yaml because there is already great software(like zappa) that does this.

Warning: aws-lambda deploy target still considered beta

bert-etl ships with a deploy target to aws-lambda. This feature isn't very well documented yet, and has quite a bit of work to de done so it may function more consistently. Be aware that aws-lambda is a product ran and controlled by AWS. If you incure charges using bert-etl while utilizing aws-lambda, you may not consider us responsible. bert-etl is offered under MIT license which includes a Use at your own risk clause.

Begin with

Lets begin with an example of loading data from a file-server and than loading it into numpy arrays

$ virtualenv -p $(which python3) env
$ source env/bin/activate
$ pip install bert-etl
$ pip install librosa # for demo project
$ docker run -p 6379:6379 -d redis # bert-etl runs on redis to share data across CPUs
$ bert-runner.py -n demo
$ PYTHONPATH='.' bert-runner.py -m demo -j sync_sounds -f

Release Notes

0.3.0

• Added Error Management. When an error occurs, bert-runner will log the error and re-run the job. If the same error happens often enough, the job will be aborted

0.2.1

• Added Release Notes

0.2.0

• Added Redis Service auto run. Using docker, redis will be pulled and started in the background
• Added Redis Service channels, sometimes you'll want to run to etl-jobs on the same machine

Fund Bounty Target Upgrades

Bert provides a boiler plate framework that'll allow one to write concurrent ETL code using Pythons' microprocessing module. One function starts the process, piping data into a Redis backend that'll then be consumed by the next function. The queues are respectfully named for the scope of the function: Work(start) and Done(end) queue. Please consider contributing to Bert Bounty Targets to improve this documentation

https://www.patreon.com/jbcurtin

Roadmap

• Create configuration file, bert-etl.yaml
• Support conda venv
• Support pyenv venv
• Support dynamodb flush
• Support multipule invocations per AWS account
• Support undeploy AWS Lambda
• Support Bottle functions in AWS Lambda

Tutorial Roadmap

• Introduce Bert API
• Explain bert.binding
• Explain comm_binder
• Explain work_queue
• Explain done_queue
• Explain ologger
• Explain DEBUG and how turning it off allows for x-concurrent processes
• Show an example on how to load timeseries data, calcualte the mean, and display the final output of the mean
• Expand the example to show how to scale the application implicitly
• Show how to run locally using Redis
• Show how to run locally without Redis, using Dynamodb instead
• Show how to run remotly using AWSLambda and Dynamodb
• Talk about dynamodb and eventual consistency