[OpenAFS] ext[23] "testing" (was "Re: Unclean shutdown: AFS client ate my filesystem (ext3)")

[Derek Atkins]

Rudolph T Maceyko <rtm@cert.org> writes:

> OK, this is beginning to sound like superstition.  We've been using
> ext3 cache partitions for quite a while now with no ill effects...

Note that many superstitions do have some basis in fact....

> --On Wednesday, July 10, 2002 09:45:06 -0400 Derek Atkins
> <warlord@MIT.EDU> wrote:
> 
> > This is why the recommendation is two fold:
> >         1) use ext2.
> >         2) use a separate cache partition.
> >
> > #1 is because it's been tested and known to work
> 
> What extra "testing" has been done for ext2 that has not been done for
> ext3?

The cache manager was original written and tested against ext2.  ext3
came later, and no such testing was done to see how the cache
interacts with the journaling system.  Personally I have never looked
at the ext3 code to see whether the interfaces used by the cache
manager violate the ext3 journaling semantics or not.  There certainly
exist examples of file systems where the FS semantics are NOT met by
AFS caching (e.g. reiserfs).

> Rudy

-derek

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord@MIT.EDU                        PGP key available