Podman / Podman-Compose: How to set uid / map uid / use user namespace to make a volume accessible for a pod / image

Question

I want to run Postgres in podman using a local volume to store Postgres data. When using "podman only" the setup is flawless, when using podman-compose, the uid/gid the container is running with differs from the uid/gid the volume is created with, so access does not work.

I am on Windows 11, using WSL2. Versions are:

$ podman-compose version
podman-compose version: 1.0.7
['podman', '--version', '']
using podman version: 4.8.1
podman-compose version 1.0.7
podman --version
podman version 4.8.1

It all works fine when I use:

podman volume create pu-postgres-vol --driver=local
podman pod create --name pu-postgres-pod -p 9876:80 -p 5432:5432
podman run --name pu-delivery-order --pod=pu-postgres-pod -v pu-postgres-vol:/var/lib/postgresql/data -e POSTGRES_DB=delivery-order -e POSTGRES_HOST_AUTH_METHOD=trust -e POSTGRES_USER=A_USER_ID -e POSTGRES_PASSWORD=A_PASSWD -d docker.io/postgres:latest

... ps -ef on the WSL2 conatiner shows that postgres operates under the dynamic (?) uid 525286:

525286    2018  2016  0 10:46 ?        00:00:00 postgres
525286    2076  2018  0 10:46 ?        00:00:00 postgres: checkpointer
525286    2077  2018  0 10:46 ?        00:00:00 postgres: background writer
525286    2079  2018  0 10:46 ?        00:00:00 postgres: walwriter
525286    2080  2018  0 10:46 ?        00:00:00 postgres: autovacuum launcher
525286    2081  2018  0 10:46 ?        00:00:00 postgres: logical replication launcher

... and the directory also belongs to uid 525286:

$ ls -al /home/user/.local/share/containers/storage/volumes/pu-postgres-vol/
total 12
drwx------  3 user   user   4096 Dec 21 10:45 .
drwxrwxrwx  5 user   user   4096 Dec 21 10:45 ..
drwx------ 19 525286 525286 4096 Dec 21 10:46 **_data**

Things break when I use podman-compose, this is my compose file:

services:
  pu-delivery-order-postgres:
    container_name: pu-delivery-order-postgres
    image: postgres:latest
    environment:
      - POSTGRES_USER=${POSTGRES_PUdeliveryOrder_USER}
      - POSTGRES_PASSWORD=${POSTGRES_PUdeliveryOrder_PW}
      - POSTGRES_DB=${POSTGRES_PUdeliveryOrder_DB} 
    ports:
      - "${POSTGRES_PUdeliveryOrder_PORT}:5432"
    volumes:
      - pu-delivery-order-postgres-data:/var/lib/postgresql/data
    restart: always

# ----- VOLUMES FOR STORAGE -----
volumes:
  pu-delivery-order-postgres-data:
    driver: local
    driver_opts:
      size: "25MB"

This exact configuration results in:

Error: mounting volume podman-infrastructure-local_pu-delivery-order-postgres-data for container eefe22617dfd5f50a597822693d054c64a71f0e297be3ec12fe0891a4496f61a: mount: /home/user/.local/share/containers/storage/volumes/podman-infrastructure-local_pu-delivery-order-postgres-data/_data: permission denied.

The problem seems to be that the container uid is a dynamic one:

$ps -ef
...
525286    1393  1386  0 10:44 ?        00:00:00 postgres
525286    1462  1393  0 10:44 ?        00:00:00 postgres: checkpointer
525286    1463  1393  0 10:44 ?        00:00:00 postgres: background writer
525286    1465  1393  0 10:44 ?        00:00:00 postgres: walwriter
525286    1466  1393  0 10:44 ?        00:00:00 postgres: autovacuum launcher
525286    1467  1393  0 10:44 ?        00:00:00 postgres: logical replication launcher

... while the volume is created by docker-compose like this:

$ ls -al /home/user/.local/share/containers/storage/volumes/
total 20
drwxrwxrwx  5 user user 4096 Dec 21 12:42 .
drwx------ 10 user user 4096 Dec 21 12:42 ..
drwx------  3 user user 4096 Dec 21 12:42 podman-infrastructure-local_pu-delivery-order-postgres-data

I am new to podman and have limited knowledge about container and Linux tech. However, I understand that a mismatch of owner and group id and ownership between a process and a volume cause problems in Linux.

I tried every way I could find to play with userns mappings I could imagine:

  pu-delivery-order-postgres:
    container_name: pu-delivery-order-postgres
    image: postgres:latest
    # userns_mode: "host"
    # user: "1000:1000"
    # userns_mode: "keep-id:uid=999,gid=999"
    # gidmap: "1000:999"
    # uidmap: "1000:999"

... but could not fix the issue.

I need help to either run the container with a different uid or fix ownership of the volume. Thank you for your help!