Distributed Locks
Last Updated: May 28, 2026
A distributed lock gives one client exclusive access to a shared resource across many processes and machines.
An in-process mutex relies on guarantees the operating system provides. A distributed lock has none of those guarantees and has to survive crashes, partitions, process pauses, and clock drift instead.
Distributed locks are used for:
- ensuring only one worker runs a scheduled job
- serializing writes to a shared external resource
- coordinating migrations or maintenance across a fleet
- single-flight protection in front of a cache or downstream service
The lock store decides who currently holds the lock. Everyone else waits, retries, or skips the work.