On 5/1/15 2:09 PM, Nick Tan wrote: > On Friday, May 1, 2015, Andrew Gabriel <[email protected]> wrote: > > However since zfs is CoW it should be ok on subsequent rsyncs.
Watch out! SMR disks are *in principle* CoW and ZFS is *in principle* CoW, but that doesn't mean that they are the same kind of CoW (indeed, I can tell you for a fact that they aren't). For starters, SMR drives come in several flavors. There's drive-internal SMR, where the drive transparently handles remapping of blocks & rewrites, and then there's host-aware SMR, where the host OS needs to explicitly use a specialized SCSI command set to query and manage SMR zones manually. I figure you've got a drive-internal SMR type of thing, otherwise your pool would just plain not work. For now (and possibly the foreseeable future), ZFS and SMR just don't mix all that well. The capacity benefit of SMR has yet to outweigh the added complexity and performance limitations of integrating support for it in ZFS, so to my knowledge, nobody has done host-aware SMR yet. Maybe in the future as capacities change, somebody will go ahead and implement it. But ATM, SMR + ZFS remains a path where there be dragons! Cheers, -- Saso _______________________________________________ openindiana-discuss mailing list [email protected] http://openindiana.org/mailman/listinfo/openindiana-discuss
