[OpenAFS] vice partition sizes, and large disks

Derrick J Brashear shadow@dementia.org
Tue, 19 Jun 2001 10:35:36 -0400 (EDT)

[Please don't CC me on any replies]

On Tue, 19 Jun 2001, Andrei Maslennikov wrote:

> On Sat, 16 Jun 2001, chas williams wrote:
> > solaris' ufs logging probably does what you want.  it stores the metadata
> > operations in a logfile and periodically flushes out the log.   from the 
> > man page:
> Hi Chas, 
> we are very interested in performance issues for large /vicepX partitions
> on Solaris, and your advice and numbers on logging options sounded quite
> interesting. However, we ran in some posting which recommends to RUN AWAY
> from ufs logging for /vicepX on Solaris, as it may do a lot of harm
> (see: http://www.mail-archive.com/info-afs@transarc.com/msg06045.html).
> After reading that mail I feel I would be at least very cautious in
> enabling logging at this time. As we see that ufs logging may be quite
> advantageous, we will probably first set up a testbed server to see how it
> all works. 
> Can you comment on this?

The whole point of this discussion was the use of the namei interface
fileserver rather than the stock inode interface, and that makes all the
difference. The author of that post was using an inode interface
fileserver. When you use a namei fileserver you have essentially regular
files for everything rather than playing directly with the inodes and
hence you're pretty much just like every other user of the filesystem, and
not subject to the whims of your OS vendor as far as your use of spare
inode fields conflicting with changes to their product. 

That message is essentially out of scope for this discussion, or if
anything is a pointer to why you need a namei fileserver if you want to
use ufs logging.