We log search terms to improve our documentation. For more information, read our Privacy Policy.

User Namespaces and Podman

Last modified: 2026 November 17


Overview

User namespaces let Podman run containers as an account’s own user instead of root. Containers never run as root, and there’s no root container daemon. If a container becomes compromised, the attacker has, at most, the access of the account that owns it. User namespaces also let each container have its own resource limits.

Disabling user namespaces can be a short-term workaround for certain types of kernel vulnerabilities. However, it can cause problems on some servers.

Important:

If you disable user namespaces, everything that runs through the ea-podman script stops working on that server. This includes AI App Hosting and EasyApache 4’s container-based packages. Examples include Apache Tomcat®, Redis®, and Valkey™.

User namespaces and kernel vulnerabilities

Inside a user namespace, an account can reach kernel features that normally require root access. These include networking and filesystem handling. A bug in one of these kernel features can let any account on the server gain root access. On shared hosting, that includes anyone who compromises a hosted website. Because of this, some security advisories recommend disabling user namespaces. However, disabling them does not fix the vulnerability itself.

  • On some distributions, the only available option is to disable user namespaces for every user, including the root user. Without user namespaces, accounts cannot run podman containers. This also affects other container tools and application sandboxes.
  • CageFS does not address these vulnerabilities. It isolates each account’s files, but it does not limit which kernel features an account can reach.

On servers that use the ea-podman script, we strongly recommend the following actions:

  • Patch your kernel promptly. A patched kernel is the lasting fix for this type of vulnerability. Live patching, where available, lets you apply the fix without a reboot.
  • On some operating systems, such as Ubuntu® 24.04 and later, you can limit user namespace creation to approved programs instead of disabling it for everyone. For more information, read your operating system vendor’s hardening guidance. AlmaLinux 8 and 9 and CloudLinux™ 8 and 9 don’t offer this option.
  • Block kernel modules you don’t use. Many of these vulnerabilities are in kernel modules that a hosting server rarely needs. Blocking an unused module protects the server regardless of whether user namespaces are enabled. Before you block a module, confirm that nothing on the server depends on it. For example, ea-podman containers need the tun module, and firewalld needs the nf_tables module.
  • If you disable user namespaces, re-enable them after you patch the kernel. Record the original setting, make the change easy to find, and note why you made it. When you re-enable them, restrict them to only what your users and containers need, and block any kernel modules you don’t use. Nothing that runs through the ea-podman script works on that server in the meantime.

On servers that do not use the ea-podman script, you can disable user namespaces.

Additional Documentation