During any redis lua script (atomic/exclusive?) execution, any writes that have occured during the script, but before the error will be committed/written and not rolled back as part of the implementation.
I am just wondering about the situation where your script is mainly bullet proof (e.g preconditions checked, but the instance executing your script crashes half way, with some writes already committed and stored in AOF log (or propagated to other masters in case of the upcoming redis-cluster)? How do you recovery/what are the best practices for this?
Also I would like to double check if the redis script is executed atomically/exclusively, as in, no other operations can occur? I am fairly sure this is the case, although does this also hold true for the upcoming redis cluster implementation and would multiple masters execute scripts concurrently?