Shell Scripting: Expert Recipes for Linux, Bash, and More

Recipe 16-4: Locks

It can be useful to know that a process has exclusive access to a resource. For example, if a script takes a few minutes to process and update a critical file, leaving it in an inconsistent state until it has finished, it is not a good idea for other scripts to try to access that file while it is temporarily in a known-bad state.

A lock system provides such a mechanism, granting only one process at any time access to a particular resource. The resource itself can be anything at all; the lock is simply the barrier that stops multiple processes doing something at once. This can be used to ensure that init scripts do not start multiple copies of their processes (often using /var/run/app-name.pid as the lock file in this case), or anything else that requires guaranteed unique access.

The normal way to achieve locking in the shell is to put the PID of the running process into a common file. Any other instances of the script check that file before entering the critical section of code, and continue only if no other instance has already claimed the lock. In practice, it is not quite that simple. Processes have to allow for the possibility that the lock file has changed since they last checked its state. This cannot be fixed simply by checking the state again because the state after that is also unknown.

Although this recipe is fine for managing a resource among two ...