You have a cluster of identical servers sharing a database. They each have a similar list of cron tasks that they might want to execute. Each task needs to run on a single machine, so the servers need to decide which one will run each job so that each task will perform on one server. The servers have no idea how many of them are in the cluster at any given moment. Suggest a distributed arbitration algorithm similar to all machines that have no single point of failure (besides the database if it uses it) to perform that arbitration task.