db-operator

This operator can manage databases, based on kubernetes definitions. It does expect the Server/Cluster to exist already. So it will not and probably will never create an RDS instance. What it will do for you:

• DB Backups
• DB Copies
• DB Restores
• Users and Privileges

Why an operator? Can't I just do that in Terraform?

The goal of this operator is to let developers manage their own databases from their own Repository. This works well with a GitOps approach. The first goal why I started writing these operators is that you can also do Preview Databases on feature branches.

Design

Databases Diagram

Backup Restore Diagram

Backup stages diagram

This repo uses the original images of postgres or msql to build it's backups. For CockroachDB we use the internal functionality

Examples / Kuttl tests

The Kuttl tests are quite good examples of how to implement a feature. You'll have to ignore the assertions of course

Postgres	CockroachDB	MySQL
backup job	backup job	backup job
backup cron job	backup cron job	backup cron job
restore job		restore job
restore cron job		restore cron job
copy job		copy job
copy cron job		copy cron job

Privileges

Examples from Kuttl tests:

• Cockroach privileges of migration user + application user with least privilege access
• MySQL privileges example on how to give SELECT on a table

Privileges are split into:

• Server privileges
• DB Privileges
  • DB Scoped
  • Schema Scoped (not implemented for mysql)
  • Table Scoped
  • Default privileges (postgres / cockroachdb) This is required if you want access to tables created in the future

Postgres / CockroachDB privileges

Scoped To	Possible Privileges
Server	SUPERUSER, CREATEROLE, CREATEDB, INHERIT, LOGIN, REPLICATION, BYPASSRLS
Database	CREATE, CONNECT, TEMPORARY (pg only), TEMP (pg only), BACKUP (crdb only), RESTORE (crdb only), ALL (temp not for cockroachdb)
Table	SELECT, INSERT, UPDATE, DELETE, TRUNCATE, REFERENCES, TRIGGER, BACKUP (crdb only), ALL
Schema	CREATE, USAGE

DbPrivs Examples:

Database:

- scope: example-db
  privs: CONNECT
  priv_type: database

Schema:

- scope: example-db.schema1
  privs: USAGE
  grantor_user_name: migration-user
  priv_type: schema

Tables:

- scope: example-db
  privs: "/table1:select,delete"
  priv_type: table

DefaultPrivs example:

- scope: example-db.TABLES
  privs: SELECT,INSERT,UPDATE,DELETE
  grantor_user_name: migration-user
  priv_type: defaultTable

Dev Requirements

• docker
• kind
• kuttl
• golang

Running tests

Quick way:

docker pull postgres:latest
make docker-build
make kind-cluster
make deploy-test-infra
make deploy-kind
make kuttl-test

Manually:

docker pull postgres:latest
make kind-cluster
make deploy-test-infra
make install
# set up port forwards
# see Running the operator on your machine for dns entries in your hosts file
make kuttl-test-cockroachdb-debugmode
make kuttl-test-mysql-debugmode
make kuttl-test-postgres-debugmode

Building / Packaging

# up helm chart version in helm/charts/db-operator/Chart.yaml
# git commit
make docker-buildx
make generate-deploys
# git add new tgz file
# git commit and push

Creating new controllers

operator-sdk create api --group db-operator --version v1alpha1 --kind <KIND> --resource --controller

Running the operator on your machine with the resources in Kind cluster

add this line to your /etc/hosts file:

127.0.0.1	localhost postgres.postgres.svc.cluster.local mysql.mysql.svc.cluster.local cockroachdb-public cockroachdb-public.cockroachdb.svc.cluster.local

make start-test-cluster
kubectl -n postgres port-forward svc/postgres 5432 &

If you want to run as binary

make run

Vscode:

{
    "name": "Debug",
    "type": "go",
    "request": "launch",
    "mode": "debug",
    "program": "${workspaceRoot}"
}

If you want to run the test scenario's while you're in debug mode:

make deploy-test-yamls-postgres

cleanup:

kind delete cluster