From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stefan Priebe Subject: Re: osd dies / crahes directly after mkcephfs Date: Fri, 15 Jun 2012 23:50:04 +0200 Message-ID: <4FDBAE0C.6010606@profihost.ag> References: <4FDBA78C.4050401@profihost.ag> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from mail.profihost.ag ([85.158.179.208]:58766 "EHLO mail.profihost.ag" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757323Ab2FOVuD (ORCPT ); Fri, 15 Jun 2012 17:50:03 -0400 In-Reply-To: Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Sage Weil Cc: "ceph-devel@vger.kernel.org" Am 15.06.2012 23:46, schrieb Sage Weil: > On Fri, 15 Jun 2012, Stefan Priebe wrote: >> Hi, >> >> i've seen several osd crashes on one of my machines directly after creating >> the ceph fs. >> >> Attached is the osd log. I also have a core dump file. Do you need it? > > Also, are you mounting the ceph file system on the same node as the osd? > This might be a sync(2)-induced deadlock. No i use only rados / rbd and in this case i have never used the fs at all as it has happened directly after mkcephfs and -a start. So i didn't even had the chance. > Putting a ceph fs mount on the same node as an osd is dicey in low memory > situations. On older glibcs and kernels without syncfs(2) support, it can > easily deadlock. Even this does not apply i have syncfs. my send log shows: mount syncfs(2) syscall fully supported (by glibc and kernel) Stefan