From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ipmail07.adl2.internode.on.net ([150.101.137.131]:25567 "EHLO ipmail07.adl2.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751781AbdLCWEq (ORCPT ); Sun, 3 Dec 2017 17:04:46 -0500 Date: Mon, 4 Dec 2017 09:03:59 +1100 From: Dave Chinner Subject: Re: Filesystem crash Message-ID: <20171203220359.GW5858@dastard> References: <3f4d2b55-9dd1-bc5a-2e49-cdfdc9a134b3@genome.arizona.edu> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3f4d2b55-9dd1-bc5a-2e49-cdfdc9a134b3@genome.arizona.edu> Sender: linux-xfs-owner@vger.kernel.org List-ID: List-Id: xfs To: Chandler Cc: linux-xfs@vger.kernel.org On Fri, Dec 01, 2017 at 03:35:07PM -0700, Chandler wrote: > Hi, we had our filesystem crash the other day after getting 100% > full (although 53GB were still reported free). I had to reboot the > system and use xfs_repair. Yup, it tripped over an freespace tree corruption and shut down to prevent it from being propagatd. Could have been caused by anything - hardware, kernel memory corruption, a bug, a MD rebuild issue, etc. Did you save the output of xfs_repair so we can see what errors it fixed up? > It seem to me this shouldn't happen just > because it got full, so maybe there is some other issue? The > filesystem resides on an MD RAID-5 array with 4x 2TB disks that are > in good health, the array gets checked weekly by mdadm. The only > error messages in the system log were related to XFS (see below). > The OS is RHEL 6.9 with kernel 2.6.32-642.11.1.el6.x86_64. Not much we can do to diagnose the problem on old RHEL kernels on upstream lists - the codebase has diverged from upstream too far. Report the problem to your local RHEL support engineer if you need further diagnostic help. Cheers. Dave. -- Dave Chinner david@fromorbit.com