On Sun, Aug 07, 2005 at 11:19:12PM -0500, Colin Ingram wrote:
> calling all lvm/raid gurus....
No guru, but have just suffered through something similar. I just built
an lvm on top of raid10 and certainly went through all of your issues
below
>
> I'm running debian sarge and I have/had a 200GB raid 1 devices with LVM2
> on top. The raid one was made from hdb1 and hdc1. Recently when I was
> trying to change the size of a logical volume. I noticed that my vg was
> using /dev/hdc1 as a physical volume instead of /dev/md0.
before we go further, I had to do two tweaks to the /etc/lvm/lvm.conf.
First, you need to make sure you've got:
md_component_detection = 1
Then, I needed to add a filter to remove all of my /dev/hd* drives so
that lvm would only be allowed to see the /dev/md* "drives". There was
already a line there to remove /dev/cdrom, so I hijacked that:
filter = [ "r|/dev/cdrom|", "r|/dev/hd.*|" ]
In the end, I'm left with the following in my /etc/lvm/.cache
# This file is automatically maintained by lvm.
persistent_filter_cache {
valid_devices=[
"/dev/md5",
"/dev/md0",
"/dev/md1"
]
}
md0 is raid1 /, md1 is raid1 swap, and md5 is raid10 over md2/3/4
That resolved my found duplicate PV issues
>
> $ pvscan
> Found duplicate PV YDMoRSv4EKZXHNPuchid5hPIwmXDCBCm: using
> /dev/hdb1 not /dev/hdc1
> PV /dev/hda4 VG LVM lvm2 [68.96 GB / 0 free]
> PV /dev/hdc1 VG data lvm2 [186.30 GB / 4.00 MB free]
> Total: 2 [255.26 GB] / in use: 2 [255.26 GB] / in no VG: 0 [0 ]
>
> some additional digging and I found that my raid devices is inactive
>
> $ mdadm -D /dev/md0
> mdadm: md device /dev/md0 does not appear to be active.
>
> but the mdsuperblock still seems present on both devices
>
> $mdadm -E /dev/hdb1
> /dev/hdb1:
> Magic : a92b4efc
> Version : 00.90.00
> UUID : 9dc1df9b:771959ca:53ed751d:7b3b40ba
> Creation Time : Sun Apr 17 13:24:05 2005
> Raid Level : raid1
> Raid Devices : 2
> Total Devices : 2
> Preferred Minor : 0
>
> Update Time : Fri May 13 16:52:37 2005
> State : clean
> Active Devices : 2
> Working Devices : 2
> Failed Devices : 0
> Spare Devices : 0
> Checksum : 1dcab098 - correct
> Events : 0.482221
>
>
> Number Major Minor RaidDevice State
> this 1 3 65 1 active sync
> /dev/.static/dev/hdb1
>
> 0 0 22 1 0 active sync
> /dev/.static/dev/hdc1
> 1 1 3 65 1 active sync
> /dev/.static/dev/hdb1
>
> $ mdadm /dev/hdc1
> /dev/hdc1:
> Magic : a92b4efc
> Version : 00.90.00
> UUID : 9dc1df9b:771959ca:53ed751d:7b3b40ba
> Creation Time : Sun Apr 17 13:24:05 2005
> Raid Level : raid1
> Raid Devices : 2
> Total Devices : 2
> Preferred Minor : 0
>
> Update Time : Fri May 13 16:52:37 2005
> State : clean
> Active Devices : 2
> Working Devices : 2
> Failed Devices : 0
> Spare Devices : 0
> Checksum : 1dcab069 - correct
> Events : 0.482221
>
>
> Number Major Minor RaidDevice State
> this 0 22 1 0 active sync
> /dev/.static/dev/hdc1
>
> 0 0 22 1 0 active sync
> /dev/.static/dev/hdc1
> 1 1 3 65 1 active sync
> /dev/.static/dev/hdb1
>
>
> I'm not sure how this happen but I've come up with a couple of
> possibilities that happen several months ago.
>
> 1. I migrated these disk from a previous test box for this server. I
> may have messed on the initial install into the new box.
> 2. I replaced hdb1 as hdc1 when doing some maintenance.
>
> In any case LVM recognized the UUID of one of disks and continued
> normally.
>
> Now I'm try to fix the situation. All the data on the disks are backup
> so I could just start fresh but I would like to repair the situation if
> I can. Here is the steps I've come up with. Please check them out for
> pitfalls etc or missing commands.
>
> Outline:
> To fix the LVM, I will treat the situation as if I am migrating the
> disks to a new system. Instead of moving drives I will be repairing raid.
> To fix Raid I need to determine which disk LVM has been using then
> restart the RAID with that disk then add the second disk as a spare.
>
> Detail:
> 1. Umount file systems
>
> # unmount /backup
> # unmount /share/ftp
> # unmount /share
>
> 2. Mark Volume Group as inactive
>
> # vgchange -an data
>
> 3. Export volume Group
>
> # vgexport data
>
> 4. Mount Disk and examine contents for recent data
>
> # mount /dev/hdb1 /mnt/hdb
> # mount /dev/hdc1 /mnt/hdc
>
> 5. Restart array with devices with most recent data
>
> # mdadm -A /dev/md0 /dev/hdb1 or /dev/hdc1
> (will this start the array degraded or do I need to create a new array
> with mdadm -C /dev/md0 --level=1 --raid-devices=/dev/hdb1 missing)
>
> 6. Add second devices as spare
>
> # mdadm /dev/md0 --add /dev/hdc1 or /dev/hdb1
>
> 7. Fix LVM (this is assuming that pvscan now recognizes PV /dev/md0 as
> being apart of VG data
>
> # vgimport data
> # vgchange -ay data
> # mount /share
> # mount /share/ftp
> # mount /backup
>
> 8. Set up mdadm in monitor mode and have a beer!
>
>
> Thanks for the help in advance
>
> Colin
>
>
> --
> To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> with a subject of "unsubscribe". Trouble? Contact
> [EMAIL PROTECTED]
>
>
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
