From mboxrd@z Thu Jan 1 00:00:00 1970 From: Phil Turmel Subject: Re: Found duplicate PV: using /dev/sda3 not /dev/md1 Date: Mon, 16 Jun 2014 10:35:26 -0400 Message-ID: <539F00AE.9090504@turmel.org> References: <1896515726.20140614235853@oudeis.org> <1454723160.20140615203503@oudeis.org> <539E5A90.3080207@turmel.org> <19411862.20140616155327@oudeis.org> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: <19411862.20140616155327@oudeis.org> Sender: linux-raid-owner@vger.kernel.org To: =?UTF-8?B?UmFpbmVyIEbDvGdlbnN0ZWlu?= Cc: linux-raid List-Id: linux-raid.ids Hi Rainer, [CC to list added back. Use reply-to-all on kernel.org lists] On 06/16/2014 09:53 AM, Rainer F=C3=BCgenstein wrote: >=20 > PT> Your problem is that you are using version 0.90 metadata. It and= v1.0 > PT> put the superblock at the end of the member devices, and it canno= t be > PT> found if the device size changes. Plus, the data starts at secto= r zero, > PT> so if the MD superblock isn't found, the device content is identi= fied > PT> without the raid layer. >=20 > If I understand it correctly, the error happened (apart from the > superblock version) because of the reboot(s) before a) md1 was grown > and/or/maybe b) sync was finished. Probably, but v0.9 arrays are vulnerable to whole-disk vs. partition misidentification, and should never be used on a partition that encompasses the end of a disk. > PT> These metadata formats should only be used with boot partitions t= hat > PT> need to be readable without assembling the array. > this was how the centos installer created the raid (with 0.90). will > keep it that way on md0 (which holds /boot and root). >=20 > PT> You need to stop /dev/md1 as soon as possible to eliminate the ch= ances > PT> of further corruption. >=20 > [root@gateway home]# mdadm --manage --stop /dev/md1 > mdadm: fail to stop array /dev/md1: Device or resource busy > Perhaps a running process, mounted filesystem or active volume group? >=20 > [root@gateway home]# lsof | grep md1 > md1_raid1 461 root cwd DIR 9,0 4096 = 2 / > md1_raid1 461 root rtd DIR 9,0 4096 = 2 / > md1_raid1 461 root txt unknown = /proc/461/exe >=20 > since nothing else than md1_raid1 is accessing /dev/md1, I assume it > is safe to --force --stop? What is running as process ID 461? That needs to be killed off, and an= y mount involving /dev/md1 stopped. > PT> If you must keep the server online, I recommend the following ste= ps: >=20 > big thanks, your strategy is much better than the one I had planned > (and fills some knowledge gaps). Your situation is the primary reason I recommend LVM. Phil -- To unsubscribe from this list: send the line "unsubscribe linux-raid" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html