GPT on Question 14, Sep 25

The usual Bash pattern is to track child PIDs and clean them up from an EXIT trap:

#!/usr/bin/env bash

children=()

cleanup() {
    rc=$?
    trap - EXIT HUP INT TERM

    if ((${#children[@]})); then
        kill -TERM -- "${children[@]}" 2>/dev/null || true
        for pid in "${children[@]}"; do
            wait "$pid" 2>/dev/null || true
        done
    fi

    exit "$rc"
}

trap cleanup EXIT
trap 'exit 129' HUP
trap 'exit 130' INT
trap 'exit 143' TERM

worker_one &
children+=("$!")

worker_two &
children+=("$!")

wait "${children[@]}"

This handles normal exit and catchable signals. It does not run if the parent is killed with SIGKILL, the machine crashes, etc.

If children may create grandchildren, PID tracking is insufficient. On Linux, start each workload in its own process group/session and kill the group:

groups=()

setsid worker_one &
groups+=("$!")

cleanup() {
    rc=$?
    trap - EXIT HUP INT TERM

    for pgid in "${groups[@]}"; do
        kill -TERM -- "-$pgid" 2>/dev/null || true
    done
    wait 2>/dev/null || true

    exit "$rc"
}

Avoid kill 0: it kills the shell’s entire process group and may include the invoking shell or unrelated processes.

For cleanup that must also work after an uncatchable parent death, Bash alone cannot guarantee it. Use one of:

• •a service manager/cgroup, such as systemd with KillMode=control-group
• •a small launcher using Linux prctl(PR_SET_PDEATHSIG, SIGTERM)
• •children that monitor a pipe from the parent and exit on EOF

For production process-tree management, a service manager/cgroup is generally the most reliable option.