Bash – using flock in bash, why does killing a child process kill the parent process too

bash

In the code snippet below, I want the script to be limited to only running one copy at a time, and for it to restart server.x if it dies for any reason.

Without flock involved, the loop correctly restarts if I kill the server process, but once I use flock to ensure the script is only running once, if I kill server.x it also kills the parent process.

How can I ensure that killing the child process in a flock script keeps the parent around?

#!/bin/bash
set -e
(
    flock -x -n 200
    while true
    do
        ./server.x $1
    done
) 200>/var/lock/.m_rst.$1.lock

Best Answer

I think that the lock file might be bothering the subprocess, causing it to terminate with a non-zero exit status: set -e will cause your script to terminate if one of the commands inside exit non zero for example:

#!/bin/bash
set -e
echo hello
false
echo goodbye

Outputs:

hello

It's also worth noting that the flock(2) manpage indicates that the flock system call will copy the file handle to forked processes and across execve; this might mean that server.x is unhappy having the lock file open. In the following example the lock will be held by both processes:

#!/bin/bash
(
  flock -x -n 200
  bash
) 200>/tmp/test.lock

Executing this test while performing lsof | grep /tmp/test.lock will show the lockfile is opened by both processes.

Related Topic