Update instance (blue/green deployment)

https://www.claudiokuenzler.com/blog/1240/haproxy-how-to-disable-enable-backend-server-command-line-cli-socket

• Open a shell in the HAProxy docker container

docker exec -it k4proxy /bin/sh

• Inside the container

# Make sure, socat is installed
apk add socat

# show backend servers
echo "show servers state" | socat unix-connect:/var/run/haproxy.sock stdio

# disable one server
echo "disable server k4webservers/s1" | socat unix-connect:/var/run/haproxy.sock stdio

• Update instance (outside the container)

./deploy2PROD1.sh

• If everything is fine (Logs, etc.), then enable the backend server in the docker container again

# enable first server again
echo "enable server k4webservers/s1" | socat unix-connect:/var/run/haproxy.sock stdio

# and disable the other server
echo "disable server k4webservers/s2" | socat unix-connect:/var/run/haproxy.sock stdio

• Then update the other instance (outside the container)

./deploy2PROD2.sh

Upgrade PostGres (different DB install users, at least one of them "root"; i.e. before v16)

• Dump old data

docker exec -it pdgbK4P /bin/bash
pg_dump -d k4photographers -U root > dumpK4P.sql

• stop all containers
• copy /docker/db directory to /docker/db<oldVersion>
• empty /docker/db/data directory (keep docker-entrypoint-initdb.d)
• start (upgraded) postgres container pgdbK4P to initialize the new version database
• Restore DB dump on new machine (dwhprovider data restore will not work)

docker exec -it pdgbK4P /bin/bash
psql -d k4photographers -U postgres < dumpK4P.sql

• Start K4P Grails Application to initialize everything code related (or additional missing columns)
• Run etl.groovy script with MIN_INVOICE=1 and MIN_ASSET=1 to fetch all data initially.

Upgrade PostGres (same DB install users, but not "root"; i.e. from v16 on)

Background: The pg_upgrade command is designed to upgrade f.e. the layout of the system tables; while a simple dump-and-restore writes all data again and might fail with constraint errors, pg_upgrade aims to update system data only and to reuse the old user data as-is. Unfortunately the "root" table is a system table, which is also present for a database install user "root". So, if you chose to install PostGres against the recommendations as "root" it breaks pg_upgrade (and probably pg_restore etc., too).

• stop all containers
• copy /docker/db directory to /docker/db<oldVersion> (and additional backup folder)
• empty /docker/db directory, (re)add empty /docker/db/data folder (no docker-entrypoint-initdb.d!)
• start (upgraded) postgres container pgdbK4P to initialize the new version database
• stop the (upgraded) postgres container again
• build the upgrade container

docker build -t pg_upgrade_15-16:latest -f Upgrade_Dockerfile .

# Upgrade_Dockerfile
FROM postgres:16

ENV PGBINOLD=/usr/lib/postgresql/15/bin/
ENV PGBINNEW=/usr/lib/postgresql/16/bin/
ENV PGDATAOLD=/var/lib/postgresql/olddata
ENV PGDATANEW=/var/lib/postgresql/data

RUN apt-get update && apt-get install -y postgresql-15

USER postgres
WORKDIR /var/lib/postgresql/
CMD [ "/usr/lib/postgresql/16/bin/pg_upgrade" ]

• start the upgrade container with both db volumes mounted

docker run -ti -v "./db15/data:/var/lib/postgresql/olddata" \
               -v "./db/data:/var/lib/postgresql/data" pg_upgrade_15-16