From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sage Weil Subject: Re: cmon: PGMonitor::encode_pending() assert failure Date: Thu, 3 Feb 2011 14:24:43 -0800 (PST) Message-ID: References: <20110203220245.GA6179@onthe.net.au> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Return-path: Received: from cobra.newdream.net ([66.33.216.30]:36519 "EHLO cobra.newdream.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751540Ab1BCWXz (ORCPT ); Thu, 3 Feb 2011 17:23:55 -0500 In-Reply-To: <20110203220245.GA6179@onthe.net.au> Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Chris Dunlop Cc: ceph-devel@vger.kernel.org On Fri, 4 Feb 2011, Chris Dunlop wrote: > On Thu, Feb 03, 2011 at 01:03:17PM -0800, Sage Weil wrote: > > Hi Chris, > > > > This is an interesting one. Would it be possible for you to > > tar up your mondata directory on the failed node and post it > > somewhere I can get at it? From the looks of things the pgmap > > incremental state file is truncated, but I'd like to confirm. > > > > http://tracker.newdream.net/issues/762 > > Aw crap, sorry, I blew that fs away installing the latest master > to see what happened there ...whereupon overnight I've promptly > hit the "WARNING: at fs/btrfs/inode.c:2143" problem*. > > I can revert back to my previous install and run the same > workload to see if it crops up again if that's useful (it took > about 12 hours of rsync'ing files into the fs to get there), or > I can try the workload using latest ceph with Josef Bacik's > btrfs-work** to see if either problem crops up again. Any > preference? Let's go with latest master and latest bits from Josef. It's the future! Thanks! sage