pg-repack-docker

Docker image for 'pg_repack' (PostgreSQL extension) to use 'client-side' to run/invoke the actual repack functionality for a PostgreSQL database with the extension installed.

Reference: https://github.com/reorg/pg_repack

This has been modified to use the version of postgres we're using in prod and to match the version of pg_repack available on RDS.

The following commands depend on having the version you want to use and tag in the following env var:

TAG_VERSION=1.4.8

Build

docker buildx build --platform linux/x86_64 -t pg_repack:$TAG_VERSION .

Push to ECR

aws ecr get-login-password --region us-west-2 | docker login --username AWS --password-stdin 537761503111.dkr.ecr.us-west-2.amazonaws.com
docker tag pg_repack:$TAG_VERSION 537761503111.dkr.ecr.us-west-2.amazonaws.com/pg_repack:$TAG_VERSION
docker push 537761503111.dkr.ecr.us-west-2.amazonaws.com/pg_repack:$TAG_VERSION

Run

# All tables:

kubectl run pgrepack --env="PGPASSWORD=<password>" --image=537761503111.dkr.ecr.us-west-2.amazonaws.com/pg_repack:1.4.8 --restart=Never pg_repack -- -h <db-url> -p 5432 -U affinity --dbname=affinity --no-superuser-check -j 4

# single table

kubectl run pgrepack --env="PGPASSWORD=<password>" --image=537761503111.dkr.ecr.us-west-2.amazonaws.com/pg_repack:1.4.8 --restart=Never pg_repack -- -h <db-url> -p 5432 -U affinity --dbname=affinity --no-superuser-check -j 4 --table <tableName>    

Monitoring

You can follow progress of the repack using kubectl logs -f pgrepack If you need to stop the repack for whatever reason (space is getting too low for example), you can kill the pod using kubectl delete pod pgrepack and it will usually clean up after itself. If it didn't, you can cleanup by removing the pg_repack extension and then reinstalling, that forces it to clean itself up.