Migration from the KQ Implementation

Jason J. W. Williams jasonjwwilliams at gmail.com
Wed May 4 21:21:54 EDT 2011


> No, that's just basic journal recovery and no fsck is needed. ext[3,4] does
> that, and there is still fsck for when that just isn't sufficient.

Actually it's not journal recovery because there is no journal. There
is no replay of journal entries, it's removing the most recent
incomplete/bad writes from the pool.

> You are still only talking about missing commits. I'm talking about
> different scenarios that aren't based on just the recent data getting
> trashed.

The recent data is what you care about to get the pool imported. Once
it's imported you can use a scrub to rebuild the bad blocks if there's
sufficient redundancy. If the corruption is random throughout the
structure on all disks, you're going to run into problems traversing
the block tree structure fsck tool or not.

-J



More information about the zfs-discuss mailing list