Processes and zombie fixes

The Processes screen refreshes every 2 seconds and lists each process with its PID, parent, user, state, CPU, memory, run time and command, plus the top 10 by CPU and by memory.

Signals

ActionSignalEffect
Ask it to collect zombiesSIGCHLDHarmless reminder to a parent process to collect its finished children. Does not stop anything.
EndSIGTERMAsks the process to shut down; it finishes open work and exits. A service may restart it.
KillSIGKILLStops it at once; unsaved work is lost. Try ending it gracefully first.

Before any signal WECOMMANDER checks that the process name still matches (a PID reused by another program is refused). Processes that keep the server reachable are refused, and a database needs an extra confirmation because stopping it can lose data being written.

Zombie processes

A zombie has already finished and uses no CPU or memory, but it stays in the process table until its parent collects it; it cannot be killed directly. The Zombie processes view groups them by parent and shows the Docker container or systemd service that owns the parent.

  1. First send the parent the harmless reminder (SIGCHLD).
  2. If the program does not collect them, restart its service or container. It is down for a few seconds; the zombies go with the old process.
  3. For a permanent fix, run the container with a small init process (--init) that collects finished processes. For apps deployed with Coolify, “Apply suggestion” adds --init to the app's custom Docker options through Coolify's API and redeploys it; a Coolify API key with write and deploy permissions is asked once and kept in this computer's keychain.

SSH, Docker, containerd, journald, logind, D-Bus and WECOMMANDER itself are never restarted.

For AI

server_processes is read-only. server_process_signal (reap, term, kill) and server_restart_owner need a human approval for that exact call unless the device or SSH server is on full trust; the same protections apply.