Date: Mon, 2 Nov 2009 09:28:47 +0100 From: Borja Marcos <borjam@sarenet.es> To: Pawel Jakub Dawidek <pjd@FreeBSD.org> Cc: freebsd-fs@freebsd.org, stef@memberwebs.com Subject: Re: 8.0-RC2: ZFS deadlock with zfs receive Message-ID: <BCB5F6B2-A84F-4D40-9C8E-FF15AADAF62B@sarenet.es> In-Reply-To: <20091030232805.GA2996@garage.freebsd.pl> References: <804B79F6-27CE-40D2-8AB8-6FC378F448FA@sarenet.es> <4AEA0EAD.1050302@memberwebs.com> <20091030232805.GA2996@garage.freebsd.pl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Oct 31, 2009, at 12:28 AM, Pawel Jakub Dawidek wrote: > On Thu, Oct 29, 2009 at 03:52:45PM -0600, Stef Walter wrote: >> Borja Marcos wrote: >>> I've been sending some alltraces to pjd about this easy to reproduce >>> problem. I am using zfs send/zfs receive to replicate a dataset >>> from one >>> server to another. At 1 minute intervals, an incremental snapshot is >>> sent to update the dataset copy. If there is reading activity on the >>> dataset copy, a deadlock can happen rendering ZFS and all the FS >>> subsystem unusable. I've tried with 8.0RC2 and it still happens. > > I was able to reproduce it, but I don't have fix yet. > >> FWIW, another (or the same) zfs recv deadlock I've been trying to >> get to >> the bottom of: >> >> http://lists.freebsd.org/pipermail/freebsd-fs/2009-October/ >> 006999.html > > Could you guys recompile your kernel after uncommenting line: > > #CFLAGS+=-DDEBUG=1 > > in sys/modules/zfs/Makefile? > > You should see panic on assertion instead of deadlock, I believe. Aye aye, Sir! What do you want me to collect? Just an alltrace? Anything else? If you want the VMWare images to examine the panic by yourself just let me know. Borja.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BCB5F6B2-A84F-4D40-9C8E-FF15AADAF62B>