A library to help run a highly-scalable AWS service that performs data processing tasks in response to external events. You provide the messages and the logic to process them, while Watchbot handles making sure that your processing task is run at least once for each message. Watchbot is similar in many regards to AWS Lambda, but is more configurable, more focused on data processing, and not subject to several of Lambda's limitations.
Install and use the current watchbot version
Add these lines to your Dockerfile, to use the latest watchbot for the linux operating system.
RUN wget https://s3.amazonaws.com/watchbot-binaries/linux/v7.0.0/watchbot -O /usr/local/bin/watchbot
RUN chmod +x /usr/local/bin/watchbot
os: You can replace
linux with other operating systems like
tag: You can replace
v7.0.0 with any watchbot tag starting from and more recent than v4.0.0
If you are an existing user of watchbot, take a look at "Upgrading to Watchbot 4", for a complete set of instructions to upgrade your stacks to Watchbot 4.
- queue: An SQS queue is a "backlog" of messages for your stack to process that helps to guarantee every message gets processed at least once. It can be a standard queue or a first-in-first-out (FIFO) queue.
- message: A message in the queue represents some job to be processed. Generally, you are responsible for sending messages to your stack by publishing to Watchbot's SNS topic, or optionally by POST to a webhook endpoint (see
WatchbotUseWebhooks in the parameters section below).
- worker: CLI command/subprocess responsible for processing a single message. You define the work performed by the worker through the
command property in the cloudformation template. Watchbot sets environment variables for the subprocess that represent the content of a single message.
- watcher: The main process in the container that polls the queue, spawns worker subprocesses to process messages, and tracks results.
- task: The ECS task that contains both the watcher main process and all of the worker subprocesses. You specify how many tasks to run using the
minSize parameters in the cloudformation template. The service scales the number of tasks based on the existence of messages in the queue.
- cluster: An ECS cluster of EC2s that are used to run the watcher and worker containers that make up your service.
- template: A CloudFormation template defining the resources Watchbot needs
What you provide:
- a cluster that Watchbot will run on
- a docker image representing your task, housed in an ECR repository and tagged with a git sha or a git tag
- a CloudFormation template defining any configuration Parameters, Resources, and Outputs that your service needs in order to perform its processing.
💡 Other prerequisites:
cloudformation-kms-production deployed according to the instructions in cloudformation-kms. Makes encryption of sensitive environment variables that need to be passed to ECS simple using cfn-config.
What Watchbot provides:
- a queue that triggers your workers
- an SNS topic through which you'll send messages to the queue, unless you have a FIFO queue
- an ECS TaskDefinition for your worker, using the image you provide
- one or more watcher containers that run continuously on your cluster, polling the queue, running a worker for each message, removing messages from the queue as workers complete, and managing worker failures and retries
- a script to help you include the resources Watchbot needs to run in your template
Building a Watchbot service
- Create a Github repository for your code.
- Write and test the code that a worker will perform in response to a message.
- Write a Dockerfile at the root of your repository which defines the steps required to bootstrap a worker. Rather than specifying a
CMD instruction, use the
command property in the cloudformation template to indicate what will be executed when your worker is launched in response to a message. Note that message details will be provided as environment variables to your worker subprocess, and that your worker's exit code will determine whether the message is deleted or returned to the queue (see below).
- Use the Dockerfile to build an image and store it in an ECR repository. See ecs-conex for a CI framework to do this for you whenever you commit to the repository.
- Write and deploy your service using a CloudFormation template that watchbot helps you to build.
- Building a Watchbot template
- The worker's runtime environment
- What happens when workers fail?
- Logging and metrics
- Using Watchbot's reduce-mode
- Watchbot's command-line utilities
- Watchbot's CloudWatch alarms