Skip to content
Discussion options

You must be logged in to vote

I am not sure primarycache=none is a very good idea, it is too synthetic. But flushing the caches by exporting/importing the pool before the test would make sense. And the test should be short enough to complete before all metadata will be cached. With default indirect block size of 128KB with ~1024 pointers per indirect block you need to read only about 0.1% of data randomly to already cache all the metadata in ARC.

You'd probably need more of different metadata types, such as tons of small files, directories, xattrs, etc to make a bigger benefits from special vdev. Also special vdev is almost mandatory in case regular vdevs unable to store small blocks, such as DRAID, or may be some RAI…

Replies: 2 comments 2 replies

Comment options

You must be logged in to vote
1 reply
@yuanfangcc
Comment options

Comment options

You must be logged in to vote
1 reply
@yuanfangcc
Comment options

Answer selected by yuanfangcc
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Category
Q&A
Labels
None yet
3 participants