Concept

Each facility has a database server. For each site the facility library is the main library. The libraries from other facilities are replications. These replications are read-only. This means that every site is independent but still benefits from the work and libraries another facility is doing.

In Postgres terms:
A logical replication is the preferred way to go, since each facility can have their own database setup and also different version of Postgres without running into any issues.

That’s not possible Physical replication, since the Postgres version as well as the folder/file structure needs to be identical. With logical we have one Publisher and as many Subscribers as we like. The Subscribers will be the Postgres databases in the other facilities.

Requirements

In each facility you have a database server up and running with Postgres installed.

Monitoring

To monitor the databases you can for example use pgwatch2

Setup VMs

install Postgres

Configure Logical Replication

1. initiate 2 new PostgresSQL database clusters

2. Configure Publisher server with “wal_level=logical“

3. Start the instance

4. Create a database and the tables

Configure primary server

# create database cluster
initdb -D /var/lib/postgresql/daselement-munich_publication

Edit the config to set the primary database cluster to the Logical Replication mode

# edit postgresql.conf
vi /var/lib/postgresql/daselement-munich_publication/postgresql.conf

# edit line for wal_level to logical
wal_level = logical

Now start the database cluster

# start the database cluster
pg_ctl -D /var/lib/postgresql/daselement-munich_publication start

Create Postgres Users - make sure that the Users exist in all the other databases as well

# access the database via psql
psql --port=5432 postgres

# create superuser for IT/admins and pipeline TDs
CREATE ROLE postgres LOGIN password 'postgres';
ALTER USER root CREATEDB;
ALTER USER root CREATEROLE;
ALTER USER root SUPERUSER;
ALTER USER root REPLICATION;

# create user that is allowed to create new databases and read/write information
CREATE ROLE user_daselement LOGIN password 'password';
ALTER USER root CREATEDB;

# create replication user that is used by the other facilites
CREATE ROLE user_replication WITH REPLICATION LOGIN password 'password';

# exit the database from psql
\q

Move to a supervisor/lead artists workstation.

Now create a new library inside the software. Use the dbuser to connect to the databas.

Hurray, we have the main library for facility Munich running.

The next step is create a publication for this newly create database. Back to the Postgres server.

# log in to the newly created database
psql --port=5432 de_munich_vfxelements

We have to create a publication for this library/database that all the other facilities can later subscribe to.

# create publication for all tables
create publication muc_pub for all tables;

# the expected output is: CREATE PUBLICATION

Configure subscriber server

1. Create subscriber database cluster

2. copy database Schema from the primary database

3. create subscription to primary database

1 create database

Now let’s setup a database in a different facility

# create database cluster
initdb -D /var/lib/postgresql/daselement-munich_subscription

# start the database cluster
pg_ctl -D /var/lib/postgresql/daselement-munich_subscription start

2 copy schema

Postgres replication does not support the Postgres Data Definition Language (DLL) commands.
For us this means that we have to copy over the Database Schema from the Primary Database to the subscriber server. This step only needs to be done once.

# create database from primary database
psql --port=5460 postgres

# create an empty database where the information will be replicated to
create database rep_de_munich_vfxelements

# expected ouptut: CREATE DATABASE

Now copy over all the information from the primary to the subscriber. Execute this command on the primary database server.

# use pg_dump to copy the data
pg_dump -s de_munich_vfxelements -p 5432 | psql -h 192.168.178.50 -p 5432 rep_de_munich_vfxelements

3 create subscription

The last step is now to subscribe to the primary facility publication.

dbname = name of the database in the munich facility (muc_pub)

host = ip-of-publisher-postgres-server (e.g. the munich postgres server)

user = must be a superuser or replication role

create subscription muc_subscription connection 'dbname=de_munich_vfxelements host=192.168.178.50 user=user_replication port=5433' publication muc_pub;

Configure Logical Replication

1. Create a publication on the Publish Server

2. Create a subscription on the Subscriber Server

3. Test the replication

Troubleshooting

userful command

psql commands

initdb: command not found

The initdb is located in the installation directory of Postgres.

Fix for Ubuntu:

# create a symbolic link to initdb to access the command
sudo ln -s /usr/lib/postgresql/14/bin/initdb /usr/local/bin

# now you can run ...
initdb

pg_ctl: command not found

The pg_ctl is located in the installation directory of Postgres.

Fix for Ubuntu:

# create a symbolic link to initdb to access the command
sudo ln -s /usr/lib/postgresql/14/bin/pg_ctl /usr/local/bin

# now you can run ...
pg_ctl

Unable to connecto to server; FATAL: role “postgres“ does not exist

Seems like there is no user in the database called “postgres”.
To create a user use this command:

# access the database
psql --port=5432 postgres

# create user in the database
CREATE ROLE postgres LOGIN password 'postgres';

START STREAMING

This is for a streaming replication! We want to have a logical replication.

Create primary database server

initdb -D /var/lib/postgresql/daselement-munich

Configure primary server

1. change postgres.conf to enable networking

2. create a replication user

3. allow remote access in pg_hba.conf

change postgres.conf to enable networking

vim /var/lib/postgresql/daselement-munich/postgres.conf

# uncomment and change the line:
listen_addresses = '*'

# and change the port to 5433 to avoid conflicts with production port 5432
port = 5433

# start the Postgres server
pg_ctl -D /var/lib/postgresql/daselement-data start

create a replication user

# create replication user
psql --port 5433 postgres

[local]:5433 postgres=# create user repuser replication;

allow remote access in pg_hba.conf

add line to allow access for the replication user

vim /var/lib/postgresql/daselement-data/pg_hba.conf

# TYPE  DATABASE        USER            CIDR-ADDRESS            METHOD
# IPv4 local connections
host    all             repuser             localhost               trust

# restart the Postgres server
pg_ctl -D /var/lib/postgresql/daselement-munich restart

The primary system is now ready. Let’s move on to the replica system!

Create replica database server

# create the replica Postgres server
# this command pg_basebackup copies the files from the primary database

pg_basebackup -h db-munich -U repuser --checkpoint=fast -D /var/lib/postgresql/daselement-munich-replica/ -R --slot=daselement_munich -C --port 5433

In the replica directory you will find these two files

standby.signal

this file defines that the database that it is a replica

postgresql.auto.conf

contains information to connect to the primary server

Now start the replica server

# start the Postgres server
pg_ctl -D /var/lib/postgresql/daselement-munich-replica

END STREAMING