Animl Ingest

Lambda function for ingesting and processing camera trap images.

About

The animl-ingest stack is a collection of AWS resources managed by the Serverless framework. When users or applications such as animl-base upload images (or .zip files of images) to the animl-staging-<stage> bucket, one or more Lambda functions are fired and perform the following:

if a .zip file is detected (i.e. a user initiated a bulk upload from the animl-frontend user interface), the file is unzipped, the contents are validated, ML processing resources are dynamically spun up to process the batch, and the images are copied into the ingestion bucket
when new images are detected in the ingestion bucket, a separate lamnda (ingest-image) fires, which:
extracts EXIF metadata
creats a thumbnail of the image
stores the thumbnail and the original in buckets for production access
passes along the metadata in a POST request to a graphQL server to create a record of the image metadata in a database
deletes the image from the staging bucket

Setup

Prerequisits

The instructions below assume you have the following tools globally installed:

Serverless
Docker
aws-cli

Create "animl" AWS config profile

The name of the profile must be "animl", because that's what serverles.yml will be looking for. Good instructions here.

Make a project direcory and clone this repo

git clone https://github.com/tnc-ca-geo/animl-ingest.git
cd animl-ingest
npm install

IMPORTANT NOTE: Sharp, one of the dependencies that's essential for opening image files and performing inference, may have issues once deployed to Lambda. See this GitHub Issue and their documentation for more info on deploying Sharp to Lambda, but in short, unless your node_modules/@img directory has the following binaries, the inference Lambda will throw errors indicating it can't open Sharp:

sharp-darwin-arm64
sharp-libvips-darwin-arm64
sharp-libvips-linux-x64
sharp-libvips-linuxmusl-x64
sharp-linux-x64
sharp-linuxmusl-x64

Running the following may help install the additional necessary binaries if they are not present, but it seems a bit inconsistent:

npm install --os=linux --cpu=x64 sharp

Dev deployment

From project root folder (where serverless.yml lives), run the following to deploy or update the stack:

# Deploy or update a development stack:
serverless deploy --stage dev

# Deploy or update a production stack:
serverless deploy --stage prod

Prod deployment

Use caution when deploying to production, as the application involves multiple stacks (animl-ingest, animl-api, animl-frontend), and often the deployments need to be synchronized. For major deployments to prod in which there are breaking changes that affect the other components of the stack, follow these steps:

Set the frontend IN_MAINTENANCE_MODE to true (in animl-frontend/src/config.js), deploy to prod, then invalidate its cloudfront cache. This will temporarily prevent users from interacting with the frontend (editing labels, bulk uploading images, etc.) while the rest of the updates are being deployed.
Manually check batch logs and the DB to make sure there aren't any fresh uploads that are in progress but haven't yet been fully unzipped. In the DB, those batches would have a created: property but wouldn't yet have uploadComplete or processingStart or ingestionComplete fields. See this issue more info: https://github.com/tnc-ca-geo/animl-api/issues/186
Set ingest-image's IN_MAINTENANCE_MODE to true (in animl-ingest/ingest-image/task.js) and deploy to prod. While in maintenance mode, any images from wireless cameras that happen to get sent to the ingestion bucket will be routed instead to the animl-images-parkinglot-prod bucket so that Animl isn't trying to process new images while the updates are being deployed.
Wait for messages in ALL SQS queues to wind down to zero (i.e., if there's currently a bulk upload job being processed, wait for it to finish).
Backup prod DB by running npm run export-db-prod from the animl-api project root.
Deploy animl-api to prod.
Turn off IN_MAINTENANCE_MODE in animl-frontend and animl-ingest, and deploy both to prod, and clear cloudfront cache.
Copy any images that happened to land in animl-images-parkinglot-prod while the stacks were being deployed to animl-images-ingestion-prod, and then delete them from the parking lot bucket.

Related repos

Animl is comprised of a number of microservices, most of which are managed in their own repositories.

tnc-ca-geo / animl-ingest

readme