Date: Wed, 11 Dec 2002 15:27:14 -1000 From: Clifton Royston <cliftonr@lava.net> To: "Ronald G. Minnich" <rminnich@lanl.gov> Cc: Joshua Goodall <joshua@roughtrade.net>, Michael Grant <mg-fbsd3@grant.org>, freebsd-cluster@FreeBSD.ORG Subject: Re: sharing files within a cluster Message-ID: <20021211152714.A25854@lava.net> In-Reply-To: <Pine.LNX.4.44.0212101530270.1460-100000@carotid.ccs.lanl.gov>; from rminnich@lanl.gov on Tue, Dec 10, 2002 at 03:30:55PM -0700 References: <20021210222210.GG98967@roughtrade.net> <Pine.LNX.4.44.0212101530270.1460-100000@carotid.ccs.lanl.gov>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Dec 10, 2002 at 03:30:55PM -0700, Ronald G. Minnich wrote:
> On Wed, 11 Dec 2002, Joshua Goodall wrote:
>
> > * NFS probably is robust enough for some environments. A suggestion
> > I was given to remember with cross-mounts is to use soft/interruptible, and
The most critical thing with NFS cross-mounts is that all servers
involved need to have their boot sequences carefully checked for
dependencies, and preferably should have two separate phases - one
which gets the servers up to the point that they can act reliably as a
*server* for all NFS volumes they export, and only start trying to mount NFS volumes from other
servers after the first phase is completed successfully,
If you make a mistake in this - or if your default rc scripts do -
you can end up configuring a situation where your entire network is
unbootable if more than one server goes down, because you have NFS
volumes listed in multiple servers' fstabs, and the servers then
deadlock trying to mount each other at startup. (This happened to me
some years back at a previous job where a lot of random cross-mounts
were set up on various HP-UX servers.)
> > always login as root directly (rather than trying to sudo from a user
> > account that may have just become unavailable in /home).
>
> soft/interruptible: bad. Leads to large blocks of zeros in files.
Interruptible is quite different from soft.
Soft = bad ("emulates a broken hard drive".)
Interruptible = GOOD! (avoids creating processes that are effectively
unkillable until the NFS server comes back up.) Note that this doesn't
lead to misread/miswritten files, unless that could normally result
from a process being killed with an interrupt signal.
man mount_nfs
...
-i Make the mount interruptible, which implies that file system
calls that are delayed due to an unresponsive server will fail
with EINTR when a termination signal is posted for the process.
...
> Spongy:
> better.
I've seen this referred to, but had not seen a system that actually
implements spongy NFS mounts. Does FreeBSD-stable or -current actually
support spongy mounts? Have I missed this in the docs?
-- Clifton
--
Clifton Royston -- LavaNet Systems Architect -- cliftonr@lava.net
"If you ride fast enough, the Specialist can't catch you."
"What's the Specialist?" Samantha says.
"The Specialist wears a hat," says the babysitter. "The hat makes noises."
She doesn't say anything else.
Kelly Link, _The Specialist's Hat_
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-cluster" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20021211152714.A25854>