Talk: SaaS on Rails on PostgreSQL

.gitignore

@@ -52,3 +52,5 @@ docker/.pgpass
 output.log
 
 .pgpass
+
+docs/posette/docker-pg-data-dir/

docs/posette/README.md

@@ -0,0 +1,45 @@
+# README
+
+## Citus
+
+## Sharding types
+As of 12.1, we have two types:
+
+- Row-based sharding
+- Schema-based sharding
+
+## General Terminology and Concepts
+- Coordinator node (the primary node)
+- Shard nodes are Postgres instances that we don't want to directly modify
+
+## Workflow
+- Reference tables or distributed tables
+- For row-based, call `create_distributed_table()` to create a distributed table
+
+## Ruby gems
+- [activerecord-multi-tenant](https://github.com/citusdata/activerecord-multi-tenant) adds helpers
+
+
+## Active Record Horizontal Sharding
+- Shard per tenant
+- ShardRecord base class
+
+```rb
+class ShardRecord < ApplicationRecord
+  self.abstract_class = true
+
+  connects_to shards: {
+    company_one: { writing: :primary_company_one, reading: :primary_company_one_replica }
+  }
+end
+```
+
+## Citus Schema Sharding
+
+```sql
+SELECT * FROM citus_schemas;
+```
+
+## citus_stat_tenants
+
+- [citus_stat_tenants](https://www.citusdata.com/blog/2023/05/12/tenant-monitoring-in-citus-and-postgres-with-citus-stat-tenants/)

docs/posette/active_record_horizontal_sharding_example.yml

@@ -0,0 +1,11 @@
+production:
+  primary:
+    database: my_primary_database
+  primary_replica:
+    database: my_primary_database
+    replica: true
+  primary_company_one:
+    database: company_one_db
+  primary_company_one_replica:
+    database: company_one_db
+    replica: true

docs/posette/citus_examples.sql

@@ -0,0 +1,105 @@
+-- Generic example
+-- Multi-tenant applications
+---
+-- https://docs.citusdata.com/en/v12.1/get_started/tutorial_multi_tenant.html
+
+-- Add tables: companies, campaigns, ads
+-- "Distribution column" is the company ID
+-- This is either a PK or a FK
+-- companies.id
+-- campaigns.company_id
+-- ads.company_id
+
+-- Then distribute the tables:
+-- SELECT create_distributed_table('companies', 'id');
+-- SELECT create_distributed_table('campaigns', 'company_id');
+-- SELECT create_distributed_table('ads', 'company_id');
+
+-- Tables are "colocated"
+
+-- Bulk load data using COPY command
+-- \copy companies from 'companies.csv' with csv
+-- \copy campaigns from 'campaigns.csv' with csv
+-- \copy ads from 'ads.csv' with csv
+
+--docker exec -it citus psql -U owner -d rideshare_development
+
+-- Rideshare examples
+
+CREATE TABLE companies (
+    id BIGINT GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
+    name text NOT NULL,
+    image_url text,
+    created_at timestamp without time zone NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updated_at timestamp without time zone NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+-- `rideshare.users` could be a reference table
+-- `rideshare.locations` could be a reference table
+-- Needed to drop FKs referring to the table I wanted to make a reference table
+-- alter table trip_requests drop constraint fk_rails_3fdebbfaca;
+-- alter table trip_requests drop constraint fk_rails_fa2679b626;
+-- alter table trip_requests drop constraint fk_rails_c17a139554;
+-- alter table trips drop constraint fk_rails_e7560abc33;
+SELECT create_reference_table('users');
+SELECT create_reference_table('locations');
+
+INSERT INTO companies (name, image_url)
+VALUES ('MSP Rides', 'https://media.istockphoto.com/id/151520574/photo/white-formal-gloved-uniformed-hand-opening-car-door.jpg?s=1024x1024&w=is&k=20&c=Sg--aZrhlv4LOt6sOHEFhg548_Y2wYESnSAl-8RQGGQ=');
+
+ALTER TABLE trips ADD COLUMN IF NOT EXISTS company_id BIGINT;
+ALTER TABLE trip_requests ADD COLUMN IF NOT EXISTS company_id BIGINT;
+
+ALTER TABLE trip_requests ADD CONSTRAINT fk_trip_requests_company_id
+ FOREIGN KEY (company_id) REFERENCES companies(id);
+
+ALTER TABLE trips ADD CONSTRAINT fk_trips_company_id
+ FOREIGN KEY (company_id) REFERENCES companies(id);
+
+-- Let's associate trips with a company
+UPDATE trips
+SET company_id = (SELECT id FROM companies LIMIT 1);
+
+UPDATE trip_requests
+SET company_id = (SELECT id FROM companies LIMIT 1);
+
+\q
+
+-- Run as superuser
+docker exec -it citus psql -U postgres
+
+-- Include rideshare search path
+-- Should be SET ROLE postgres
+-- Connected to postgres DB
+SET search_path = "$user", public, rideshare;
+
+-- Row distribution
+-- ERROR:  connection to the remote node owner@localhost:5432 failed with the following error: FATAL:  too many connections for role "owner"
+-- Was 10
+-- Increase connections for owner
+SELECT create_distributed_table('companies', 'id');
+
+-- alter table trip_requests drop constraint trip_requests_pkey CASCADE;
+SELECT create_distributed_table('trip_requests', 'company_id');
+
+-- drop constraint from trip_positions, for demo: fk_rails_9688ac8706
+-- Although trip_positions would be great to be distributed
+-- Drop single column PK
+SELECT create_distributed_table('trips', 'company_id');
+
+-- Log all statements on citus Postgres:
+-- ALTER DATABASE rideshare_development SET log_statement = 'all';
+
+
+-- "Reference" tables concept
+-- These are distributed to all workers
+
+
+-- Schema-based sharding
+-- https://docs.citusdata.com/en/stable/develop/reference_ddl.html
+-- Schema distribution
+SELECT citus_schema_distribute('user_service');
+
+-- DDL propagation
+-- https://docs.citusdata.com/en/stable/develop/api_guc.html
+-- citus.enable_ddl_propagation (boolean)

docs/posette/configure_citus_docker.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+#
+# Single node citus:
+# https://docs.citusdata.com/en/v12.1/installation/single_node_docker.html
+#
+# export DOCKER_CLI_HINTS=false
+#
+docker run \
+  --name citus \
+  --publish 15001:5432 \
+  --volume ${PWD}/docker-pg-data-dir:/var/lib/postgresql/data \
+  --env POSTGRES_USER=postgres \
+  --env POSTGRES_PASSWORD=postgres \
+  --detach \
+  citusdata/citus:12.1.3
+
+# Wait a moment
+sleep 2
+
+# verify it's running, and that Citus is installed:
+PGPASSWORD=postgres psql -U postgres \
+  -h localhost \
+  -p 15001 \
+  -d postgres \
+  -c "SELECT * FROM citus_version();"

docs/posette/configure_citus_docker_two.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+#
+# Single node citus:
+# https://docs.citusdata.com/en/v12.1/installation/single_node_docker.html
+#
+# export DOCKER_CLI_HINTS=false
+#
+
+# Configure Rideshare:
+cd rideshare
+export DOCKER_CLI_HINTS=false
+export DB_URL="postgres://postgres:postgres@localhost:15001/postgres"
+export RIDESHARE_DB_PASSWORD="HSnDDgFtyW9fyFI"
+
+# sh db/setup.sh 2>&1 | tee -a citus.log
+# psql:db/create_database.sql:3: NOTICE:  Citus partially supports CREATE DATABASE for distributed databases
+# DETAIL:  Citus does not propagate CREATE DATABASE command to workers
+# HINT:  You can manually create a database and its extensions on workers.
+# CREATE DATABASE
+
+docker exec -it citus psql -U postgres
+
+\l
+\c rideshare_development
+\dx
+
+SELECT * FROM citus_version();

docs/posette/cpk_example.sql

@@ -0,0 +1,35 @@
+CREATE TABLE accounts (
+  id INT GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
+  name text UNIQUE NOT NULL
+);
+
+-- single column primary key
+CREATE TABLE orders (
+  id BIGINT GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
+  account_id INT NOT NULL,
+  CONSTRAINT fk_account_id
+    FOREIGN KEY (account_id)
+    REFERENCES accounts(id)
+);
+
+-- Alternative for FK
+CREATE TABLE orders (
+  id BIGINT GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
+  account_id INT NOT NULL REFERENCES accounts(id)
+);
+
+-- Drop single column PK
+ALTER TABLE orders DROP CONSTRAINT orders_pkey;
+
+-- Add CPK
+ALTER TABLE orders
+ADD CONSTRAINT orders_pkey_cpk
+PRIMARY KEY (id, account_id);
+
+-- Create with CPK
+CREATE TABLE orders (
+  id BIGINT GENERATED ALWAYS AS IDENTITY NOT NULL,
+  account_id INT NOT NULL,
+  CONSTRAINT orders_pkey_cpk
+    PRIMARY KEY (id, account_id)
+);

docs/posette/rls_config.sql

@@ -0,0 +1,58 @@
+-- NOTE: superusers override RLS
+-- Create less priviledged users to test with
+CREATE USER bob;
+CREATE USER jane;
+CREATE SCHEMA my_schema;
+GRANT USAGE ON SCHEMA my_schema TO bob;
+GRANT USAGE ON SCHEMA my_schema TO jane;
+ALTER DEFAULT PRIVILEGES IN SCHEMA my_schema
+GRANT INSERT, SELECT, UPDATE, DELETE ON TABLES TO bob;
+ALTER DEFAULT PRIVILEGES IN SCHEMA my_schema
+GRANT INSERT, SELECT, UPDATE, DELETE ON TABLES TO jane;
+
+SET search_path = 'my_schema';
+
+CREATE TABLE users (
+  user_id serial PRIMARY KEY,
+  username text UNIQUE NOT NULL
+);
+INSERT INTO users (username)
+VALUES ('bob'), ('jane');
+
+CREATE TABLE user_data (data TEXT, user_id INTEGER);
+
+CREATE OR REPLACE FUNCTION current_user_id() RETURNS int AS $$
+DECLARE
+    found_user_id int;
+BEGIN
+    SELECT user_id INTO found_user_id FROM users WHERE username = CURRENT_USER;
+    RETURN found_user_id;
+EXCEPTION WHEN NO_DATA_FOUND THEN
+    RETURN NULL; -- or raise an exception, depending on your requirements
+END;
+$$ LANGUAGE plpgsql STABLE;
+
+SET ROLE bob;
+INSERT INTO user_data (data, user_id)
+VALUES ('bob data', current_user_id());
+
+SET ROLE jane;
+INSERT INTO user_data (data, user_id)
+VALUES ('jane data', current_user_id());
+
+-- SET ROLE andy;
+-- Must be owner of table
+-- Enable for user_data
+ALTER TABLE user_data ENABLE ROW LEVEL SECURITY;
+
+-- Policy for user_data
+CREATE POLICY select_policy ON user_data
+FOR SELECT
+  USING (user_id = current_user_id());
+
+-- Make sure it's set to ON
+SET row_security TO ON;
+
+SELECT polname, polcmd, pg_get_expr(polqual, polrelid), pg_get_expr(polwithcheck, polrelid)
+FROM pg_policy
+WHERE polrelid = 'user_data'::regclass;

docs/posette/schema_examples.sql

@@ -0,0 +1,9 @@
+-- Bookstore app companies
+CREATE SCHEMA amazon;
+CREATE SCHEMA bn;
+CREATE SCHEMA bookshop;
+
+
+-- Rideshare company schemas
+CREATE SCHEMA mn_msp_rides;
+CREATE SCHEMA nyc_rides;