On Dec 29, Bruce Sterling Woodcock wrote:
> It's "wack" and it's similar to fsck for UNIX or scandisk for NT. It
> corrects any errors to the filesystem that may have been introduced
> due to unusual crashes or hardware failures.
As Bruce said, the main difference between "wack" and fsck or
scandisk is that you shouldn't need to run wack during normal
operation.
In particular, unlike UNIX and NT, you don't need to run "wack"
after a system panic, or when you pull the plug without shutting
the system down, or after a power failure. The WAFL filesystem
uses a database technique called "shadow paging with NV-RAM log
replay" to maintain a fully consistent state on disk even in the
face of unexpected shutdowns.
So when do you need to run wack? Suppose that you have a disk
fail in RAID, and then when you try to reconstruct the data from
the lost disk you discover that there is a bad sector on one of
the other drives. Most of your file system data is just fine,
but you've got a couple of blocks that can't be recovered. The
RAID subsystem zeros both blocks, since that's the best that you
can do, but now you need to run wack in order to fix things up,
in case the zeroed blocks contained file system data like
inodes, indirect blocks, or the free block map.
And yes, I must confess that like any computer, NetApp systems
do contain the occasional bug, so WAFL has a number of checks to
make sure that the on-disk data looks reasonable, and if it
finds a problem it will ask you to perform a wack.
Our goal is for customers never to need to run wack, but the
reality is that people sometimes do need to use it.
Dave
This archive was generated by hypermail 2b29 : Tue Jan 04 2000 - 10:20:55 EST