[OpenAFS] Salvager did not run automatically on solaris 9, 1.4.1-rc10

Jeffrey Hutzelman jhutz@cmu.edu
Thu, 13 Apr 2006 19:16:42 -0400


On Thursday, April 13, 2006 07:10:24 PM -0400 Dan Pritts 
<danno@internet2.edu> wrote:

> On Thu, Apr 13, 2006 at 05:49:50PM -0400, Jeffrey Hutzelman wrote:
>> I have to admit I'm a little curious why you switched from inode to
>> namei  on a Solaris server...
>
> Interestingly, I have been considering moving to solaris as my afs server
> platform (from linux) and have been wondering what the tradeoffs were
> for using the namei vs. inode fileservers.
>
> i'd always assumed that it was performance optimizations that led to the
> inode fileserver design, rather than functionality.

Yes; the original fileserver backend design was largely motivated by 
performance characteristics, with server machines that were orders of 
magnitude slower than what people have today.  Note that the namei backend 
is newer by at least 10 years, and was designed primarily to cope with 
platforms on which the inode backend was impractical.

> And i used to lie
> awake nights worrying about some Sun patch putting the OS's fsck back
> in place.

If you follow the recommended procedures, that won't be a problem, because 
you don't replace Sun's fsck.  Instead, you arrange things so that vice 
partitions use a different one, and it all works out.


> I also have been thinking that ZFS is attractive, which would rule out
> the inode server.  I doubt that is what the original poster is doing,
> though - it's a bit early in zfs's lifecycle for that.

Based on discussions with Solaris developers, I am cautiously optimistic 
that we'll be able to make something like the inode fileserver work on ZFS. 
But as you note, it's a bit early in the lifecycle for that, and not a lot 
of investigation has been done to date.

-- Jeffrey T. Hutzelman (N3NHS) <jhutz+@cmu.edu>
   Sr. Research Systems Programmer
   School of Computer Science - Research Computing Facility
   Carnegie Mellon University - Pittsburgh, PA