From mboxrd@z Thu Jan 1 00:00:00 1970 From: Edward Shishkin Subject: Re: warning: is_tree_node: node level 56362 does not match to the expected one 1 Date: Tue, 10 Jun 2008 22:11:59 +0400 Message-ID: <484EC3EF.8050909@gmail.com> References: <20080608174322.GC19640@p15145560.pureserver.info> <484C4736.3080103@gmail.com> <20080609072909.GA22951@p15145560.pureserver.info> <484DA03D.5060207@gmail.com> <20080610110119.GC22951@p15145560.pureserver.info> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20080610110119.GC22951@p15145560.pureserver.info> Sender: reiserfs-devel-owner@vger.kernel.org List-ID: Content-Type: text/plain; charset="us-ascii" To: Ralf Gross Cc: reiserfs-devel@vger.kernel.org Ralf Gross wrote: > Edward Shishkin schrieb: > >> Ralf Gross wrote: >> >>> Edward Shishkin schrieb: >>> >>>>> http://pirx.askja.de/Supermicro_Daughter_Card_Remote_Console-1.png >>>>> >>>>> With the rescue CD I could also take a look at the system logfiles >>>>> from the time of the crash: >>>>> >>>>> >>>>> >>>>> >>>> olala... it definitely seems like hardware problems.. >>>> >>>> >>> But I can't find any hardware problems for the last 2 days. >>> >>> >>> >> How did you test it? >> Were relevant controllers, cables, etc. involved? >> > > I didn't check the cables or opened the chassis. But neither memtest > nor badblocks found anything interesting. > > >> [...] >> >>>> If yes, then try to build and check the same configuration with the >>>> same hard drives on another box. No more ideas. >>>> >>>> >>> Moving the disks to another server is not possible I fear. >>> >>> >> How much space is occupied on dm-0? (df -h) >> > > ~ 1 TB. > > Anyway, I couldn't wait any longer to get the system online again. So > I started from scratch and created one new raid volume (instead of 3) > and don't use lvm anymore. As the controller has the ability to expand > raid volumes, I won't need lvm. > > Next I'll fill the filesystem and calculate the md5sums for a couple > of times. If this doesn't shows any problems, I *hope* that I can > trust the hardware. > > Any other ideas? Is there a tool that does the above thing (fill a fs > and check the md5sums x times)? > I find md5sums over fs rather useless. If rebuild-tree stops at the same block, then it would make sense to export the partition somehow to look whats up..