To counter some of the hate in this thread: I have used this to great success as an "opportunistic" locking mechanism to, for example, reduce load on a Postgres database. The winner of the race to acquire the lock would run the (expensive) query then cache the result. On lock release, the nodes waiting on the lock would then try to read the cache before trying to acquire the lock again.
Comments
To counter some of the hate in this thread: I have used this to great success as an "opportunistic" locking mechanism to, for example, reduce load on a Postgres database. The winner of the race to acquire the lock would run the (expensive) query then cache the result. On lock release, the nodes waiting on the lock would then try to read the cache before trying to acquire the lock again.