Question about JuiceFS persistency/file system level consistency capability

[pranav-srinivas]

I want to use JuiceFS to deploy 2 instances of Redis, 1 for metadata and 1 for object storage, and turn on persistence (RDB+AOF). Is there a way to guarantee file-level consistency in this scenario?

For example, a write operation to a file consists of
(1) Write to Redis#1 for metadata.
(2) Write to Redis#2 for the object of 1st block.
(3) Write to Redis#2 for the object of 2nd block.
(4) Write to Redis#2 for the object of 3rd block.

If the whole system is terminated between (3) and (4) and is then restored, we expect (1) and (2) to have to be reverted/removed in order to maintain data consistency at the file system level. This is my current understanding, please correct if wrong and explain if I am missing anything. Thanks!