Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 13 Dec 2000 08:53:25 +0000 (GMT)
From:      Terry Lambert <tlambert@primenet.com>
To:        henrich@sigbus.com (Charles Henrich)
Cc:        freebsd-fs@FreeBSD.ORG
Subject:   Re: Filesystem tuning (minimize seeks)
Message-ID:  <200012130853.BAA26821@usr08.primenet.com>
In-Reply-To: <20001212103851.C20653@sigbus.com> from "Charles Henrich" at Dec 12, 2000 10:38:51 AM

next in thread | previous in thread | raw e-mail | index | archive | help
> Im trying to tune a FreeBSD box acting as a NFS file server, and was wondering
> if anyone had any suggestions on how to handle multiple concurrent writes from
> causing the data to be written very non-sequential.  Anyone?  Thanks!

NFS writes are supposed to be sent with the file offset, the
length of the data, and the data.  The offset is dereferenced
from the file structure on the originating machine.

Is this aggregating the output of several seperate machines?

If so, there will need to be a reconning of the file offset
between the machines doing the writing, probably with an
out of band protocol of some kind.  Most commonly, this is
accomplished with token passing, where the reciever of the
token seeks to the end of the file and appends after getting
the token.

If you could provide some more information on what the client
machines are doing, and what your desired behaviour is, we
could probably give you a better approach.

If you are running commercial Linux software, and it's not
running on FreeBSD, but runs fine on Linux, then it may be
that a cached value of the file status information is being
used, and a fix needs to be applied to NFS to ensure that it
has up to date information from the server before doing the
seek (e.g. the stat that precedes the seek may not result in
wire traffic, when it should).

There are several reasons this could fail, including, but not
limited to, missing LEASE revocation operations that should
invalidate client cached state.  NFS doesn't really guarantee
distribute coherency like this, it's really a client issue.

If you are using NFS locking, well, FreeBSD does not currently
enforce this properly (though there is BSDI code that is
supposedly available, which could probably be integrated to
rectify the locking situation).


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-fs" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200012130853.BAA26821>