Need help to avoid podman to change the directory owner.

[Raul824]

After my partition corruption, I was planning to schedule backup but podman gets in the way by changing the owner of directories.

I run most pods using podman run or podman-compose.
I have set export PODMAN_USERNS=keep-id in bashrc and running podman-compose up -d command.
But it still changes the directory owner to 524357.

I have been using podman from past 4-5 months but this uid and guid mapping I can't understand.
sorry for being dumb but is there any way I can configure podman to run without changing the owner of my directories so that my Backup job can run without skipping the podman directories.

Podman version -- 5.7.1

[BitWise-0x]

The UID 524357 you're seeing is the clue here. In rootless Podman, your /etc/subuid typically allocates a subordinate range starting at 524288 with a size of 65536. So UID 524357 = 524288 + 69 — meaning container UID 69 is being mapped to host UID 524357. This tells us keep-id is not actually being applied to those containers.

Why PODMAN_USERNS=keep-id is being ignored

If you're using podman-compose, by default it creates a pod for each project. The --userns flag and --pod flag conflict in Podman — you can't set both. So PODMAN_USERNS=keep-id gets silently ignored when podman-compose runs containers inside a pod.

Fix: Disable pod creation in podman-compose

Add this to the top of your docker-compose.yml:

x-podman:
  in_pod: false

services:
  myservice:
    image: myimage
    userns_mode: "keep-id"
    volumes:
      - ./data:/data

Or run with:

podman-compose --in-pod=false up -d

With keep-id properly applied, your host UID/GID maps 1:1 into the container, so mounted directories keep your ownership.

Also check for :U volume flag

If any of your volume mounts use the :U suffix (e.g., -v ./data:/data:U), remove it. The :U flag explicitly tells Podman to recursively chown host directories to match the container user's mapped UID — which is exactly what's causing your problem.

# BAD — will chown your host directory
volumes:
  - ./data:/data:U

# GOOD — won't touch ownership
volumes:
  - ./data:/data

Reclaim already-changed directories

For directories already changed to UID 524357, you can reclaim them:

# Option A: Use podman unshare (maps back to your UID)
podman unshare chown -R 0:0 /path/to/affected/directory

# Option B: Direct chown
sudo chown -R $(id -u):$(id -g) /path/to/affected/directory

After fixing the compose config and reclaiming the directories, your backup job should work without skipping anything.