Date: Tue, 5 Sep 2006 10:53:27 +0200 From: Daniel Bond <db@danielbond.org> To: freebsd-fs@freebsd.org Subject: UFS/mksnap/softupdates deadlock issue Message-ID: <20060905085327.GA6405@spearburn.danielbond.org>
next in thread | raw e-mail | index | archive | help
Hi, anyone know the status of this? http://www.mail-archive.com/freebsd-stable@freebsd.org/msg74035.html Every system we have after RELENG_5 has a issue with mksnap (dump/fsck) on UFS2-partitions larger than say 800GB. This means we cannot use dump, which is our main backup-strategy, and we have to fsck in single-user mode if theese systems for some reason crash. This happens via ATA/SCSI-CAM/SCSI-NONCAM (like ServeRaid). I have seen many other users having issues with this, even big companies like UPS, and quite frankly I am a little suprised that it still exists in 6.1-RELEASE. How to reproduce: run dump/fsck on live filesystem, ufs2 with softupdates over 800GB (I think arround 600GB is enough, but I'm certain with 800GB). What happens: A deadlock like Tor Egge describes in his post above. The company I work for has been using FreeBSD for more than 10 years, and we are supprised how stable and easy to administrate it is. It's a really great OS and we love it, but lately we have had to turn to commersial Linux vendors (SuSE) for storage servers due to this issue. We are all using it against our will. If anyone needs a enviroment for testing/reproducing this error, I can setup our old storage server her, and give access to any freebsd commiter. If anyone knows a patch/sollution to this, please let me know, and I will report back results. I have not yet tested if this problem exists in 7-Current, but I could if anyone wants me too. -- Med vennlig hilsen / Best regards, ------------------------------------------ Daniel Bond PGP: C822C4BD ------------------------------------------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060905085327.GA6405>