Distributed lock management using conditional updates to a distributed key value data store

Abstract

A compute cluster including multiple compute nodes may implement distributed lock management using conditional updates to a distributed key value data store. It may be determined, at one or more compute nodes of a compute cluster, that particular lock is available based on a respective lock entry for the particular lock maintained in a lock manager table at a key value data store. The key value data store may be configured to perform conditional write requests for updates to data store at the key value, and may maintain data according to a distributed durability scheme. Compute nodes that determine that a lock is available may send a conditional write request to the key value data store in order to acquire the particular lock. The compute node that acquired the particular lock may be identified based on the successfully completed conditional write request to the respective lock entry.