logdna / exclusive-lock-node Goto Github PK

Create the initial package scaffolding and first version of the code

Having refresh() on a debug level is too noisy. This gets in the way with any parent apps trying to get useful debug information. Move this to a trace level.

Inspecting a lock should be able to be done by anyone. This can allow clients to clean up bad locks by inspecting useful metadata (hostname, timestamps, etc) that are stored in the lock. Also, add the TTL into the response for added context.

• Allow anyone to inspect
• Add ttl_ms to the response
• Make the response body an object

Per the initial PR, there is no guarantees that the instance refreshing the lock is the real owner: #2 (comment)

To solve this, we would need to keep state of the locks with UUIDs, etc. The problem is that we lose atomicity since redis/keydb does not have a check-and-set functionality. A lua script could be used, but that comes at a performance cost of stopping all other commands, so on something that's potentially refreshing a lot, this may be significant.