xfs
[Top] [All Lists]

best stable recent kernel?

To: xfs-oss <xfs@xxxxxxxxxxx>
Subject: best stable recent kernel?
From: Paul Anderson <pha@xxxxxxxxx>
Date: Mon, 28 Nov 2011 16:31:42 -0500
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=mime-version:sender:date:x-google-sender-auth:message-id:subject :from:to:content-type; bh=25XU8QHJNFAvP2g6dzUv0ufpxVtfa0yTcqeoiWVMVB4=; b=Np+D/kNThgIJOgtYuGuJImg2QX/b3Z6UiYX5SHiEZANkXrl3ij4aHZlBZsuWau2iMx Gbif44ZG/26lhShL74OurdB3PVm8fvd8AzhYEMJCP9spzETtcxKxlogvdnQ/fENCZ/hk RCQwoNUBOvVveJqpvzdbs0QXYZaANbhwk/6Ww=
Sender: powool@xxxxxxxxx
I need to start deploying some large XFS based fileservers servers and
wish to try a kernel newer than what we've been using (2.6.38.5 from
kernel.org).

We've stayed at 2.6.38.5 for awhile, but have had metadata loss on
many shutdowns, clean or not.  We've found 0 length files where the
files had been written and closed long before the crash - in some
cases we were careful sync by hand, wait for quiescent drives, then
clean reboot, etc.  This sounds related to a series of commits that
indicates some workarounds and fixes in the time frame of 2.6.38.5 to
3.0 (how completely this particular problem was fixed, I don't know).

Several past threads here on this list have described these servers,
but they are 126 2TB drive RAID 60 used for mostly genetic sequencing
data.

Christoph recently suggested here in another xfs developer list thread:

"I've also backported it to the 3.0-stable issues.  I would recommend
to upgrade to the latest Linux 3.0-stable release, as I'll plan to
backport all important XFS fixes to it."

Is this a decent target for us to try?  We're looking for the best XFS
and NFS stability that is practical.

Thanks,

Paul Anderson
University of Michigan

<Prev in Thread] Current Thread [Next in Thread>