Docker (compose) Postgres upgrading

Proof of concept to use a entrypoint wrapper to perform both pg_upgrade as data schema updates.

Demo

$ docker-compose -f docker-compose11.yml up -d
$ docker-compose -f docker-compose11.yml exec db psql -U postgres some_app -c "SELECT * FROM some_data;";
 id |                                         data                                          |            created            
----+---------------------------------------------------------------------------------------+-------------------------------
  1 | init data                                                                             | 2019-10-12 12:00:39.466121+00
  2 | more data                                                                             | 2019-10-12 12:00:39.466121+00
  3 | PostgreSQL 11.4 on x86_64-pc-linux-musl, compiled by gcc (Alpine 8.3.0) 8.3.0, 64-bit | 2019-10-12 12:00:39.466121+00
(3 rows)
$ docker-compose -f docker-compose12.yml up -d # and wait a bit (or remove the -d and run in separate console)
$ docker-compose -f docker-compose12.yml exec db psql -U postgres some_app -c "SELECT * FROM some_data;";
 id |                                         data                                          |            created            |         new         
----+---------------------------------------------------------------------------------------+-------------------------------+---------------------
  1 | init data                                                                             | 2019-10-12 12:00:39.466121+00 | 
  2 | more data                                                                             | 2019-10-12 12:00:39.466121+00 | 
  3 | PostgreSQL 11.4 on x86_64-pc-linux-musl, compiled by gcc (Alpine 8.3.0) 8.3.0, 64-bit | 2019-10-12 12:00:39.466121+00 | 
  4 | PostgreSQL 12.0 on x86_64-pc-linux-musl, compiled by gcc (Alpine 8.3.0) 8.3.0, 64-bit | 2019-10-12 12:01:28.735376+00 | insert in migration
(4 rows)

ToDo

• Probably copy/replace the whole docker-entrypoint.sh script.
• Run analyze_new_cluster.sh and delete_old_cluster.sh if successful
• Zero downtime migration, with logical replication?