This should be added:
GRANT ALL ON SCHEMA public TO ;

This was changed in postgresql 15 and without it, db initialization will fail.

It looks like there are a few other approaches. Looking at #168 (comment) , it is said that to create individual schema for database seems to be effective too.

Also, changing the user who executes query in create_database (value of -U option for psql command) from ${PG_USER} to ${DB_USER} may be effective. In other words, the aim is to make ${DB_USER} the owner of the database (that was suggested in sameersbn/docker-gitlab#3107 (comment)) . (In practice, a conditional branch to check whether the DB_USER parameter is set would be necessary.)

What I'm trying to say is that I'm starting to feel less confident that including that code (GRANT ...) in the kkimurak/sameersbn-postgresql codebase (kkimurak#1, kkimurak#5) was the right decision. Maybe it's something that docker-gitlab could have handled. I haven't been able to verify anything yet though..

The gitlab doc writes (https://docs.gitlab.com/install/installation/#7-database):
sudo -u postgres psql -d template1 -c "CREATE DATABASE gitlabhq_production OWNER git;"
I think in "functions" OWNER is not set. Maybe this is why official doc doesn't require the grant.