I have a (CentOS4.5) cluster in which the servers mount a GFS partition which is an LVM2 logical volume created as a mirror of two iSCSI- connected drives (with a third for the log). The LV was created using a command along the lines of: lvcreate -m 1 ... /dev/sdb /dev/sdc /dev/sdd where sd[bc] are the mirrored (iSCSI) PVs in the VG and sdd is the log. I have this working and can write data to the filesystem on one machine in the cluster and see it appear elsewhere etc. What I now want to do is to test what happens when I disable one of the mirrors and then restore it with clean disks. However, my google-foo is weak today and I can't find any information on how to get LVM to tag the broken half of the mirror as failed and then how to get it to resynchronise the mirrors once the disks are replaced. Can anyone explain how to do this? Is it even possible? James
On Mon, Jul 23, 2007 at 04:07:57PM +0100, James Fidell wrote: ...> > lvcreate -m 1 ... /dev/sdb /dev/sdc /dev/sddor use pvreate /dev/md0 (md raid1 mirror of sda/sdb/sdc)?> > where sd[bc] are the mirrored (iSCSI) PVs in the VG and sdd is the log. > I have this working and can write data to the filesystem on one machine > in the cluster and see it appear elsewhere etc. > > What I now want to do is to test what happens when I disable one > of the mirrors and then restore it with clean disks.http://mirror.centos.org/centos/4/docs/4.5/SAC_Cluster_Logical_Volume_Manager/mirrorrecover.html .02 cents Tru -- Tru Huynh (CentOS-3 i386/x86_64 Package Maintenance) http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xBEFA581B -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <http://lists.centos.org/pipermail/centos/attachments/20070723/096fe20d/attachment.sig>
Tru Huynh wrote:> On Mon, Jul 23, 2007 at 04:07:57PM +0100, James Fidell wrote: > ... >> lvcreate -m 1 ... /dev/sdb /dev/sdc /dev/sdd > > or use pvreate /dev/md0 (md raid1 mirror of sda/sdb/sdc)?AIUI, MD isn't cluster-{aware,safe} though, so I could end up with all the servers that can see the physical disks trying to do stuff with the mirrors individually, making a horrible mess? (In my configuration, all servers mount the iSCSI devices and I'm using LVM/clvmd to allow them to keep everything sane and remove any single point of failure.)>> where sd[bc] are the mirrored (iSCSI) PVs in the VG and sdd is the log. >> I have this working and can write data to the filesystem on one machine >> in the cluster and see it appear elsewhere etc. >> >> What I now want to do is to test what happens when I disable one >> of the mirrors and then restore it with clean disks. > http://mirror.centos.org/centos/4/docs/4.5/SAC_Cluster_Logical_Volume_Manager/mirrorrecover.htmlPerfect. That's exactly what I was after. Thank you. James