[Swift-devel] [Fwd: [pads-users] Mounting homes read-only on computes]

[Mihael Hategan]

You'd want that to be configurable for each site (each may have its SFS
mounted somewhere else). That would suggest it might go in the sites
file.

On Mon, 2011-04-18 at 23:55 -0500, Allan Espinosa wrote:
> The workdirectory of a site and client host maybe on different
> machines.  My guess is that it should be something in the
> etc/swift.properties file
> 
> 2011/4/18 Jonathan S Monette <jonmon at utexas.edu>:
> > What about the work directory?
> >
> > On Apr 18, 2011 9:28 PM, "Mihael Hategan" <hategan at mcs.anl.gov> wrote:
> >> It should be a configurable parameter. I was thinking more like gpfs,
> >> but it should be the user's choice.
> >>
> >> On Mon, 2011-04-18 at 18:53 -0700, Sarah Kenny wrote:
> >>> /var/tmp ?
> >>>
> >>> On Sun, Apr 17, 2011 at 1:25 AM, Mihael Hategan <hategan at mcs.anl.gov>
> >>> wrote:
> >>> I guess we should allow worker logs to go in some other place
> >>> than
> >>> ~.globus/
> >>>
> >>> Mihael
> >>>
> >>> -------- Forwarded Message --------
> >>> From: pads-users at ci.uchicago.edu
> >>> Reply-to: pads-users at ci.uchicago.edu
> >>> To: pads-users at ci.uchicago.edu
> >>> Subject: [pads-users] Mounting homes read-only on computes
> >>> Date: Sat, 16 Apr 2011 22:00:40 -0500
> >>>
> >>> -----BEGIN PGP SIGNED MESSAGE-----
> >>> Hash: SHA1
> >>>
> >>> One of the things discussed as part of the post-mortem of
> >>> today's home file server outage is mounting home directories
> >>> read-only on the compute nodes. We do this on Beagle and
> >>> haven't had any complaints for the most part. So I wanted to
> >>> solicit you to see how disruptive that might be to your
> >>> current PADS work. Homes would still be mounted read-write on
> >>> the login machines, but they wouldn't be writable on the
> >>> computes. The alternative is to push your writes to either
> >>> GPFS or the local scratch filesystems. If we don't hear any
> >>> major complaints about this, we're targeting next month's
> >>> maintenance to make this live. What are your thoughts on this?
> >>> -----BEGIN PGP SIGNATURE-----
> >>> Version: GnuPG/MacGPG2 v2.0.14 (Darwin)
> >>>
> >>> iEYEARECAAYFAk2qV9gACgkQ4RgdOxQVi0CovACfW20YM0R8uexD5PeHoX6QGZc3
> >>> LOwAnR3Iv1tokCpM81T56kSGrLNuniKg
> >>> =c5Y+
> >>> -----END PGP SIGNATURE-----
> 
> 
> 
>