[gentoo-cluster] clvm/lvm2/device-mapper

[gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Hi,

I'm quite confuse with lvm2 an clvm packages on gentoo amd64.

I have 2 xen-hosts domains0 using the same volume group (vg) including  
logical volumes (lv) for the domainsU. The physical volume is on 2 SAN  
(using raid1 lvm on logical volumes) with dual path. I use multipath- 
tools to access the right path (access are active/enabled only).

So my domains0 have 4 access on storage system (2 on each volume, one  
volume on each SAN). With multipath-tools I choose only one path per  
volume. With LVM I have created some raid1 logical volumes on the 2  
san-volumes I see on my hosts domain0. Each xen domainsU has access to  
one lvm logical volume, but I need to have these logical volumes ready  
to use on the second xen-server domain0 (for failback). So I need to  
use a locking system to use my lvm volume group on the 2 domain0, right?

If I want to install clvm, the package is masked and "flagged in read  
me" as "redondant" with lvm2. In lvm2 I can't get "clvm" flag to work  
(it is in parentheses).
My packages versions are the stables ones :
sys-fs/lvm2-2.02.36
sys-fs/multipath-tools-0.4.7-r2 (0.4.8 doesn't work)
sys-kernel/xen-sources-2.6.18-r12

Do I need clvm? What must I do to having it working?

Another question. With recent versions of lvm2, I need to uninstall  
device-mapper. How can I do it when multipath-tools depends on device- 
mapper??

Best regards

David

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Robin H. Johnson]

On Mon, Oct 19, 2009 at 03:42:16PM +0200, Ulrich David wrote:
> If I want to install clvm, the package is masked and "flagged in
> read me" as "redondant" with lvm2. In lvm2 I can't get "clvm" flag
> to work (it is in parentheses).
> My packages versions are the stables ones :
> sys-fs/lvm2-2.02.36
> sys-kernel/xen-sources-2.6.18-r12
> 
> Do I need clvm? What must I do to having it working?
There was NEVER a stable package release of sys-fs/clvm.

Correspondingly, you must use the ~arch releases of sys-fs/lvm2 for good
CLVM support.

Additionally, the clvm use flag is use.mask'd for nearly all
architectures, because nobody has tested it.
x86 is the only arch that has it unmasked:
/usr/portage/profiles/arch/x86/use.mask:-clvm

> sys-fs/multipath-tools-0.4.7-r2 (0.4.8 doesn't work)
Could you elaborate on how it doesn't work? What is your SAN hardware?

> Another question. With recent versions of lvm2, I need to uninstall
> device-mapper. How can I do it when multipath-tools depends on
> device-mapper??
Upstream LVM2 merged device-mapper into LVM2, that's why the dep has a
|| block.

-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Trustee & Infrastructure Lead
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Le 19 oct. 2009 à 22:42, Robin H. Johnson a écrit :

Hi

>> sys-fs/multipath-tools-0.4.7-r2 (0.4.8 doesn't work)
> Could you elaborate on how it doesn't work? What is your SAN hardware?

My hardware is DS4300 turbo. I have problems running kpartx from  
multipath-tools 0.4.8. It doesn't detect all my partitions... running  
on "dmsetup ls --target multipath".

>> Another question. With recent versions of lvm2, I need to uninstall
>> device-mapper. How can I do it when multipath-tools depends on
>> device-mapper??
> Upstream LVM2 merged device-mapper into LVM2, that's why the dep has a
> || block.

So, if I install lvm2 I do not need device-mapper package to be  
installed in order to use multipath-tools, right?

Thx for your answer

David

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1217 bytes --]

On Sun, Nov 01, 2009 at 08:37:55PM +0100, Ulrich David wrote:
> >>sys-fs/multipath-tools-0.4.7-r2 (0.4.8 doesn't work)
> >Could you elaborate on how it doesn't work? What is your SAN hardware?
> My hardware is DS4300 turbo. I have problems running kpartx from
> multipath-tools 0.4.8. It doesn't detect all my partitions...
> running on "dmsetup ls --target multipath".
If your drive is larger than 1TiB, there was a fix that got included
with multipath-tools-0.4.8-r1, so try that first (it's pending
stabilization).

Or sounds like this might be bug #268961 again, that I didn't have
enough info on before.
- Do you see the partitions on top of the multipath device in
  /proc/partitions?
- Can you provide a udev info dump for your multipath device?

> >Upstream LVM2 merged device-mapper into LVM2, that's why the dep has a
> >|| block.
> So, if I install lvm2 I do not need device-mapper package to be
> installed in order to use multipath-tools, right?
Correct.
LVM2 2.02.45 and newer provide device-mapper.

-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Trustee & Infrastructure Lead
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 330 bytes --]

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Le 1 nov. 2009 à 20:54, Robin H. Johnson a écrit :

> On Sun, Nov 01, 2009 at 08:37:55PM +0100, Ulrich David wrote:
>>>> sys-fs/multipath-tools-0.4.7-r2 (0.4.8 doesn't work)
>>> Could you elaborate on how it doesn't work? What is your SAN  
>>> hardware?
>> My hardware is DS4300 turbo. I have problems running kpartx from
>> multipath-tools 0.4.8. It doesn't detect all my partitions...
>> running on "dmsetup ls --target multipath".
> If your drive is larger than 1TiB, there was a fix that got included
> with multipath-tools-0.4.8-r1, so try that first (it's pending
> stabilization).

It's only 20GB volumes.

> Or sounds like this might be bug #268961 again, that I didn't have
> enough info on before.
> - Do you see the partitions on top of the multipath device in
>  /proc/partitions?

yes I see it.
The problem is on extended partitions. I have :
sda1 -> primary (boot)
sda2 -> primary (swap)
sda3 -> primary (root)
sda4 -> extended on all the rest of disk
- sda5 -> logical partition (usr)
- sda6 -> logical partition (var)

The first 3 partitions are ok (sda1 through sda3) but It takes sda4  
instead of sda5 and sda6 .... so It doesn't recognize sda5 and sda6.

> - Can you provide a udev info dump for your multipath device?

With multipath 0.4.7-r2 (working correctly) :

dom0 ~ # multipath -ll
system (3600a0b80001f3e190000552a4ab06e95) dm-2 IBM     ,1722-600
[size=20G][features=1 queue_if_no_path][hwhandler=0]
\_ round-robin 0 [prio=0][enabled]
  \_ 0:0:1:0 sdd 8:48  [active][ready]
\_ round-robin 0 [prio=3][active]
  \_ 1:0:1:0 sdh 8:112 [active][ready]
...

dom0 ~ # dmsetup ls --target multipath
system	(254, 2)
data0r1	(254, 4)
sys1domU	(254, 3)
data0r0	(254, 1)
mpath4	(254, 0)

dom0 ~ # udevadm info --query=all  --path=/sys/block/dm-2
P: /block/dm-2
N: mapper/system
L: -100
S: block/254:2
S: dm-2
S: disk/by-id/dm-name-system
S: disk/by-id/dm-uuid-3600a0b80001f3e190000552a4ab06e95
E: UDEV_LOG=3
E: DEVPATH=/block/dm-2
E: DEVNAME=/dev/mapper/system
E: DEVTYPE=disk
E: DM_NAME=system
E: DM_MAJOR=254
E: DM_MINOR=2
E: DM_STATUS=ACTIVE
E: DM_READ_ONLY=0
E: DM_EXISTS=1
E: DM_SUSPENDED=Active
E: DM_TABLE_LIVE=1
E: DM_TABLE_INACTIVE=0
E: DM_OPEN=5
E: DM_SEGMENTS=1
E: DM_EVENTS=1
E: DM_UUID=3600a0b80001f3e190000552a4ab06e95
E: DM_TARGET_TYPES=multipath
E: DEVLINKS=/dev/block/254:2 /dev/dm-2 /dev/disk/by-id/dm-name-system / 
dev/disk/by-id/dm-uuid-3600a0b80001f3e190000552a4ab06e95


dom0 ~ # udevadm info --query=all  --path=/sys/block/sda
P: /block/sda
N: sda
S: block/8:0
S: disk/by-id/scsi-3600a0b80001f3e49000055614ac5d165
E: UDEV_LOG=3
E: DEVPATH=/block/sda
E: DEVNAME=/dev/sda
E: DEVTYPE=disk
E: ID_VENDOR=IBM
E: ID_VENDOR_ENC=IBM\x20\x20\x20\x20\x20
E: ID_MODEL=1722-600
E: ID_MODEL_ENC=1722-600\x20\x20\x20\x20\x20\x20\x20\x20
E: ID_REVISION=0914
E: ID_TYPE=disk
E: ID_SERIAL=3600a0b80001f3e49000055614ac5d165
E: ID_SERIAL_SHORT=600a0b80001f3e49000055614ac5d165
E: ID_BUS=scsi
E: DEVLINKS=/dev/block/8:0 /dev/disk/by-id/ 
scsi-3600a0b80001f3e49000055614ac5d165


David

>
>>> Upstream LVM2 merged device-mapper into LVM2, that's why the dep  
>>> has a
>>> || block.
>> So, if I install lvm2 I do not need device-mapper package to be
>> installed in order to use multipath-tools, right?
> Correct.
> LVM2 2.02.45 and newer provide device-mapper.
>
> -- 
> Robin Hugh Johnson
> Gentoo Linux: Developer, Trustee & Infrastructure Lead
> E-Mail     : robbat2@gentoo.org
> GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1163 bytes --]

On Mon, Nov 02, 2009 at 01:38:56PM +0100, Ulrich David wrote:
> >Or sounds like this might be bug #268961 again, that I didn't have
> >enough info on before.
> >- Do you see the partitions on top of the multipath device in
> > /proc/partitions?
> The first 3 partitions are ok (sda1 through sda3) but It takes sda4
> instead of sda5 and sda6 .... so It doesn't recognize sda5 and sda6.
Hmm, ok, I need a little more data...

> >- Can you provide a udev info dump for your multipath device?
> With multipath 0.4.7-r2 (working correctly) :
Thanks for the 0.4.7-r2 info, just missing a few bits to make the puzzle
fit together. Below is a snippet to run, can you please provide it full
for both 0.4.7-r2 and 0.4.8-r1? Attach them in bug #268961 and CC
yourself there.

# cat /proc/partitions
# multipath -ll
# dmsetup ls --target multipath
# for i in /sys/block/* ; do udevadm info --query=all  --path=$i ; done ;
# for i in /dev/sd? ; do parted $i print ; done ;

-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Trustee & Infrastructure Lead
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 330 bytes --]

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Here it is :
http://bugs.gentoo.org/show_bug.cgi?id=268961


Le 2 nov. 2009 à 21:27, Robin H. Johnson a écrit :

> On Mon, Nov 02, 2009 at 01:38:56PM +0100, Ulrich David wrote:
>>> Or sounds like this might be bug #268961 again, that I didn't have
>>> enough info on before.
>>> - Do you see the partitions on top of the multipath device in
>>> /proc/partitions?
>> The first 3 partitions are ok (sda1 through sda3) but It takes sda4
>> instead of sda5 and sda6 .... so It doesn't recognize sda5 and sda6.
> Hmm, ok, I need a little more data...
>
>>> - Can you provide a udev info dump for your multipath device?
>> With multipath 0.4.7-r2 (working correctly) :
> Thanks for the 0.4.7-r2 info, just missing a few bits to make the  
> puzzle
> fit together. Below is a snippet to run, can you please provide it  
> full
> for both 0.4.7-r2 and 0.4.8-r1? Attach them in bug #268961 and CC
> yourself there.
>
> # cat /proc/partitions
> # multipath -ll
> # dmsetup ls --target multipath
> # for i in /sys/block/* ; do udevadm info --query=all  --path=$i ;  
> done ;
> # for i in /dev/sd? ; do parted $i print ; done ;
>
> -- 
> Robin Hugh Johnson
> Gentoo Linux: Developer, Trustee & Infrastructure Lead
> E-Mail     : robbat2@gentoo.org
> GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Just another question with dependencies... I just update lvm2 to last  
version but on my amd64 10.0 profile
"system" needs e2fsprogs and e2fsprogs-libs
lvm2 needs util-linux

but util-linux is blocked by e2fsprogs....

Is there something to do?

Regards

David


Le 4 nov. 2009 à 14:01, Ulrich David a écrit :

> Here it is :
> http://bugs.gentoo.org/show_bug.cgi?id=268961
>
>
> Le 2 nov. 2009 à 21:27, Robin H. Johnson a écrit :
>
>> On Mon, Nov 02, 2009 at 01:38:56PM +0100, Ulrich David wrote:
>>>> Or sounds like this might be bug #268961 again, that I didn't have
>>>> enough info on before.
>>>> - Do you see the partitions on top of the multipath device in
>>>> /proc/partitions?
>>> The first 3 partitions are ok (sda1 through sda3) but It takes sda4
>>> instead of sda5 and sda6 .... so It doesn't recognize sda5 and sda6.
>> Hmm, ok, I need a little more data...
>>
>>>> - Can you provide a udev info dump for your multipath device?
>>> With multipath 0.4.7-r2 (working correctly) :
>> Thanks for the 0.4.7-r2 info, just missing a few bits to make the  
>> puzzle
>> fit together. Below is a snippet to run, can you please provide it  
>> full
>> for both 0.4.7-r2 and 0.4.8-r1? Attach them in bug #268961 and CC
>> yourself there.
>>
>> # cat /proc/partitions
>> # multipath -ll
>> # dmsetup ls --target multipath
>> # for i in /sys/block/* ; do udevadm info --query=all  --path=$i ;  
>> done ;
>> # for i in /dev/sd? ; do parted $i print ; done ;
>>
>> -- 
>> Robin Hugh Johnson
>> Gentoo Linux: Developer, Trustee & Infrastructure Lead
>> E-Mail     : robbat2@gentoo.org
>> GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85
>
>

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Robin H. Johnson]

On Wed, Nov 04, 2009 at 04:33:23PM +0100, Ulrich David wrote:
> Just another question with dependencies... I just update lvm2 to
> last version but on my amd64 10.0 profile
> "system" needs e2fsprogs and e2fsprogs-libs
> lvm2 needs util-linux
> 
> but util-linux is blocked by e2fsprogs....
Upgrade your e2fsprogs+e2fsprogs-libs first.
Portage was supposed to handle that upgrade automatically.

Can we see the output of:
# emerge -pv e2fsprogs e2fsprogs-libs util-linux lvm2

Thanks for the cmd output, but just wondering if part of the 0.4.8 run
was missing something, as it jumps:
===
 254     4   12178432 dm-4
 system	(254, 0)
===

Whereas 0.4.7-r2 has:
===
 254     4    9766911 dm-4
 254     5    2411519 dm-5
dom0 ~ # multipath -ll
system (3600a0b80001f3e190000552a4ab06e95)
[size=20 GB][features=1 queue_if_no_path][hwhandler=0]
\_ round-robin 0 [prio=1][enabled]
 \_ 0:0:1:0 sda 8:0   [active][ready]
\_ round-robin 0 [prio=6][active]
 \_ 1:0:1:0 sdb 8:16  [active][ready]
dom0 ~ # dmsetup ls --target multipath
system	(254, 0)
===

-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Trustee & Infrastructure Lead
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Le 4 nov. 2009 à 20:49, Robin H. Johnson a écrit :

> On Wed, Nov 04, 2009 at 04:33:23PM +0100, Ulrich David wrote:
>> Just another question with dependencies... I just update lvm2 to
>> last version but on my amd64 10.0 profile
>> "system" needs e2fsprogs and e2fsprogs-libs
>> lvm2 needs util-linux
>>
>> but util-linux is blocked by e2fsprogs....
> Upgrade your e2fsprogs+e2fsprogs-libs first.
> Portage was supposed to handle that upgrade automatically.
>
> Can we see the output of:
> # emerge -pv e2fsprogs e2fsprogs-libs util-linux lvm2

I just add ~amd64 keywords on sys-libs/e2fsprogs-libs and e2fsprogs  
before upgrading and now It's okay. Thanks for the tip !

> Thanks for the cmd output, but just wondering if part of the 0.4.8 run
> was missing something, as it jumps:
> ===
> 254     4   12178432 dm-4
> system	(254, 0)
> ===
>
> Whereas 0.4.7-r2 has:
> ===
> 254     4    9766911 dm-4
> 254     5    2411519 dm-5
> dom0 ~ # multipath -ll
> system (3600a0b80001f3e190000552a4ab06e95)
> [size=20 GB][features=1 queue_if_no_path][hwhandler=0]
> \_ round-robin 0 [prio=1][enabled]
> \_ 0:0:1:0 sda 8:0   [active][ready]
> \_ round-robin 0 [prio=6][active]
> \_ 1:0:1:0 sdb 8:16  [active][ready]
> dom0 ~ # dmsetup ls --target multipath
> system	(254, 0)
> ===

I have put all results in the 2 files... I think the problem is on the  
extended partition (partioned with fdisk) because :
with 0.4.7-r2 :
dm-0 -> mapper/system -> (sda,sdb)
dm-1 -> mapper/system1 -> (sda1,sdb1 primary partition)
dm-2 -> mapper/system2 -> (sda2,sdb2 primary partition)
dm-3 -> mapper/system3 -> (sda3,sdb3 primary partition)
dm-4 -> mapper/system5 -> (sda5,sdb5 first part of extended partition)
dm-5 -> mapper/system6 -> (sda6,sdb6 second part of extended partition)


dom0 ~ # ls -la /dev/dm-*
lrwxrwxrwx 1 root root 13 Nov  4 14:51 /dev/dm-0 -> mapper/system
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-1 -> mapper/system1
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-2 -> mapper/system2
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-3 -> mapper/system3
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-4 -> mapper/system5
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-5 -> mapper/system6
dom0 ~ # ls -la /dev/mapper/
total 0
drwxr-xr-x  2 root root    180 Nov  4 14:51 .
drwxr-xr-x 18 root root  14260 Nov  4 13:54 ..
lrwxrwxrwx  1 root root     16 Nov  4 14:51 control -> ../device-mapper
brw-rw----  1 root disk 254, 0 Nov  4 14:51 system
brw-rw----  1 root disk 254, 1 Nov  4 14:51 system1
brw-rw----  1 root disk 254, 2 Nov  4 14:51 system2
brw-rw----  1 root disk 254, 3 Nov  4 14:51 system3
brw-rw----  1 root disk 254, 4 Nov  4 14:51 system5
brw-rw----  1 root disk 254, 5 Nov  4 14:51 system6




If I upgrade multipath to 0.4.8-r1 on a running system with mounted  
partitions (using multipath-0.4.7-r2) :



dom0 ~ # emerge -p multipath-tools
These are the packages that would be merged, in order:
Calculating dependencies... done!
[ebuild   R   ] sys-fs/multipath-tools-0.4.8-r1

dom0 ~ # emerge multipath-tools
...
dom0 ~ # dmsetup ls --target multipath --exec "/sbin/kpartx -l"
system1 : 0 1955839 /dev/mapper/system 1
system2 : 0 7815168 /dev/mapper/system 1955840
system3 : 0 7815168 /dev/mapper/system 9771008
system4 : 0 24356864 /dev/mapper/system 17586176
system5 : 0 19533823 /dev/dm-3 1
system6 : 0 4823039 /dev/dm-3 19533825

dom0 ~ # dmsetup ls --target multipath --exec "/sbin/kpartx -av"
add map system1 (254:1): 0 1955839 linear /dev/mapper/system 1
add map system2 (254:2): 0 7815168 linear /dev/mapper/system 1955840
add map system3 (254:3): 0 7815168 linear /dev/mapper/system 9771008
add map system4 (254:6): 0 24356864 linear /dev/mapper/system 17586176
device-mapper: reload ioctl failed: Invalid argument
add map system5 : 0 19533823 linear 254:6 17586177
device-mapper: reload ioctl failed: Invalid argument
add map system6 : 0 4823039 linear 254:6 37120001

dom0 ~ # ls -al /dev/mapper/
total 0
drwxr-xr-x  2 root root    200 Nov  4 21:23 .
drwxr-xr-x 18 root root  14280 Nov  4 21:23 ..
lrwxrwxrwx  1 root root     16 Nov  4 14:51 control -> ../device-mapper
brw-rw----  1 root disk 254, 0 Nov  4 14:51 system
brw-rw----  1 root disk 254, 1 Nov  4 14:51 system1
brw-rw----  1 root disk 254, 2 Nov  4 14:51 system2
brw-rw----  1 root disk 254, 3 Nov  4 14:51 system3
brw-rw----  1 root disk 254, 6 Nov  4 21:23 system4
brw-rw----  1 root disk 254, 4 Nov  4 14:51 system5
brw-rw----  1 root disk 254, 5 Nov  4 14:51 system6
dom0 ~ # ls -al /dev/dm*
lrwxrwxrwx 1 root root 13 Nov  4 14:51 /dev/dm-0 -> mapper/system
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-1 -> mapper/system1
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-2 -> mapper/system2
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-3 -> mapper/system3
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-4 -> mapper/system5
lrwxrwxrwx 1 root root 14 Nov  4 14:51 /dev/dm-5 -> mapper/system6
lrwxrwxrwx 1 root root 14 Nov  4 21:23 /dev/dm-6 -> mapper/system4


kpartx failed on mapping system4 which should be not mapped (I think,  
because its my extended base partition). The only thing which is  
changed is the addition of system4.

I hope this should help. If not I could mount a disk in my blades,  
install the system on the internal disk and try to mount a SAN volume  
on it with the same partition scheme. So I could better test it  
without boot problems.

Regards

David


> -- 
> Robin Hugh Johnson
> Gentoo Linux: Developer, Trustee & Infrastructure Lead
> E-Mail     : robbat2@gentoo.org
> GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85
>

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1236 bytes --]

On Wed, Nov 04, 2009 at 09:32:49PM +0100, Ulrich David wrote:
> If I upgrade multipath to 0.4.8-r1 on a running system with mounted
> partitions (using multipath-0.4.7-r2) :
Can you boot with 0.4.8-r1 at all? 
I'm just interested what disks turn up.

I'm trying to get some access to a SAN mounted space elsewhere to test,
so I can debug it further, should know later today if I can get access.

> kpartx failed on mapping system4 which should be not mapped (I
> think, because its my extended base partition). The only thing which
> is changed is the addition of system4.
yeah, kpartx doesn't seem to be liking the extended partition, I'm not
100% sure why yet, but it's beyond the udev rules I'm sure.

> I hope this should help. If not I could mount a disk in my blades,
> install the system on the internal disk and try to mount a SAN
> volume on it with the same partition scheme. So I could better test
> it without boot problems.
If I can't get access to the test setup I wanted, I'll get back to you
on more detailed testing.

-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Trustee & Infrastructure Lead
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 330 bytes --]

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[David Ulrich]

Le 4 nov. 2009 à 22:15, "Robin H. Johnson" <robbat2@gentoo.org> a  
écrit :

> On Wed, Nov 04, 2009 at 09:32:49PM +0100, Ulrich David wrote:
>> If I upgrade multipath to 0.4.8-r1 on a running system with mounted
>> partitions (using multipath-0.4.7-r2) :
> Can you boot with 0.4.8-r1 at all?
> I'm just interested what disks turn up.

My system is on one dual path SAN volume (sda,sdb). With 0.4.8-r1 sda1  
(/boot), sda2 (swap) and sda3 (/) are correctly mapped so I can boot.  
But sda5 (/usr) and sda6 (/var) are not mapped... Mapping breaks at  
sda4. So I don't have access to all I need (as /usr/lib for libaio,  
parted, ...). Instead sda4 is mapped which doesn't contain any  
filesystem (base extended partition).


>
> I'm trying to get some access to a SAN mounted space elsewhere to  
> test,
> so I can debug it further, should know later today if I can get  
> access.
>
>> kpartx failed on mapping system4 which should be not mapped (I
>> think, because its my extended base partition). The only thing which
>> is changed is the addition of system4.
> yeah, kpartx doesn't seem to be liking the extended partition, I'm not
> 100% sure why yet, but it's beyond the udev rules I'm sure.
>
>> I hope this should help. If not I could mount a disk in my blades,
>> install the system on the internal disk and try to mount a SAN
>> volume on it with the same partition scheme. So I could better test
>> it without boot problems.
> If I can't get access to the test setup I wanted, I'll get back to you
> on more detailed testing.

If you cannot get access to a SAN I will see if I have time to do a  
setup at work.

David


>
> -- 
> Robin Hugh Johnson
> Gentoo Linux: Developer, Trustee & Infrastructure Lead
> E-Mail     : robbat2@gentoo.org
> GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 1534 bytes --]

On Wed, Nov 04, 2009 at 11:15:59PM +0100, David Ulrich wrote:
> >I'm trying to get some access to a SAN mounted space elsewhere to
> >test, so I can debug it further, should know later today if I can get
> >access.
Ok, I've reproduced it now. kpartx is picking the wrong starting device and
offset for extended partitions. Looks like it's starting from sda4 but calling
it dm-2.

# sfdisk -qL -d /dev/sda 
# partition table of /dev/sda
unit: sectors
/dev/sda1 : start=      128, size=   524288, Id=83, bootable
/dev/sda2 : start=   524416, size= 10485760, Id=83
/dev/sda3 : start= 11010176, size=1453828654, Id= 5
/dev/sda4 : start=        0, size=        0, Id= 0
/dev/sda5 : start= 11010239, size= 50331713, Id=82
/dev/sda6 : start= 61342015, size=1403496815, Id=8e

# kpartx  -l /dev/sda
sda1 : 0 524288 /dev/sda 128
sda2 : 0 10485760 /dev/sda 524416
sda3 : 0 1453828654 /dev/sda 11010176
sda5 : 0 50331713 /dev/dm-2 63
sda6 : 0 1403496815 /dev/dm-2 50331839

# cat /proc/partitions
major minor  #blocks  name

   8     0  732421120 sda
   8     1     262144 sda1
   8     2    5242880 sda2
   8     3          1 sda3
   8     5   25165856 sda5
   8     6  701748407 sda6
 253     0    5242880 dm-0
 253     1    5242880 dm-1
 253     2   20971520 dm-2
 253     3    5242880 dm-3
 253     4   52428800 dm-4



-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Trustee & Infrastructure Lead
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 330 bytes --]

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Robin H. Johnson]

[-- Attachment #1: Type: text/plain, Size: 739 bytes --]

On Thu, Nov 05, 2009 at 12:57:06AM +0000, Robin H. Johnson wrote:
> On Wed, Nov 04, 2009 at 11:15:59PM +0100, David Ulrich wrote:
> > >I'm trying to get some access to a SAN mounted space elsewhere to
> > >test, so I can debug it further, should know later today if I can get
> > >access.
> Ok, I've reproduced it now. kpartx is picking the wrong starting device and
> offset for extended partitions. Looks like it's starting from sda4 but calling
> it dm-2.
multipath-tools-0.4.8-r2 should work for you, please test it and report
back on the bug.

-- 
Robin Hugh Johnson
Gentoo Linux: Developer, Trustee & Infrastructure Lead
E-Mail     : robbat2@gentoo.org
GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

[-- Attachment #2: Type: application/pgp-signature, Size: 330 bytes --]

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Le 5 nov. 2009 à 05:22, Robin H. Johnson a écrit :

> On Thu, Nov 05, 2009 at 12:57:06AM +0000, Robin H. Johnson wrote:
>> On Wed, Nov 04, 2009 at 11:15:59PM +0100, David Ulrich wrote:
>>>> I'm trying to get some access to a SAN mounted space elsewhere to
>>>> test, so I can debug it further, should know later today if I can  
>>>> get
>>>> access.
>> Ok, I've reproduced it now. kpartx is picking the wrong starting  
>> device and
>> offset for extended partitions. Looks like it's starting from sda4  
>> but calling
>> it dm-2.
> multipath-tools-0.4.8-r2 should work for you, please test it and  
> report
> back on the bug.

Great it works very well !

Thank you

David


>
> -- 
> Robin Hugh Johnson
> Gentoo Linux: Developer, Trustee & Infrastructure Lead
> E-Mail     : robbat2@gentoo.org
> GnuPG FP   : 11AC BA4F 4778 E3F6 E4ED  F38E B27B 944E 3488 4E85

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

Hi,

Le 19 oct. 2009 à 22:42, Robin H. Johnson a écrit :

> On Mon, Oct 19, 2009 at 03:42:16PM +0200, Ulrich David wrote:
>> If I want to install clvm, the package is masked and "flagged in
>> read me" as "redondant" with lvm2. In lvm2 I can't get "clvm" flag
>> to work (it is in parentheses).
>> My packages versions are the stables ones :
>> sys-fs/lvm2-2.02.36
>> sys-kernel/xen-sources-2.6.18-r12
>> 
>> Do I need clvm? What must I do to having it working?
> There was NEVER a stable package release of sys-fs/clvm.
> 
> Correspondingly, you must use the ~arch releases of sys-fs/lvm2 for good
> CLVM support.
> 
> Additionally, the clvm use flag is use.mask'd for nearly all
> architectures, because nobody has tested it.
> x86 is the only arch that has it unmasked:
> /usr/portage/profiles/arch/x86/use.mask:-clvm

I just need to rebound on my initial question. I now want to install lvm2 with "cluster aware" tools (clvm). I want to use a lock on my lvm volumes which could be mount by multiple hosts (but only one at a time).
I unmask clvm in the profile file of amd64. But when I install lvm2 with clvm use flag I don't have a useable clvm :

iceage0 ~ # /etc/init.d/clvmd start
 * Caching service dependencies ...
 *  Can't find service 'cman' needed by 'clvmd';  continuing...                                                                                                                                  [ ok ]
cman
 * ERROR:  Some services needed are missing.  Run
 *         './clvmd broken' for a list of those
 *         services.  clvmd was not started.
iceage0 ~ # /etc/init.d/clvmd status
 * status:  stopped

If I unmask cman, the result is :


iceage0 ~ # emerge -pv lvm2

These are the packages that would be merged, in order:

Calculating dependencies... done!

emerge: there are no ebuilds to satisfy "=sys-cluster/cman-2*".
(dependency required by "sys-fs/lvm2-2.02.51-r2" [ebuild])
(dependency required by "lvm2" [argument])

no cman-2 in portage...

I have noticed that with clvm flag, configuration option for lvm2 are :
 * econf: updating LVM2.2.02.51/autoconf/config.sub with /usr/share/gnuconfig/config.sub
./configure --prefix=/usr --build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu --mandir=/usr/share/man --infodir=/usr/share/info --datadir=/usr/share --sysconfdir=/etc --localstatedir=/var/lib --enable-readline --disable-selinux --enable-pkgconfig --libdir=/usr/lib64 --enable-dmeventd --enable-cmdlib --enable-applib --enable-fsadm --enable-static_link --with-mirrors=internal --with-snapshots=internal --with-lvm1=internal --with-cluster=internal --with-clvmd=none --with-pool=internal --sbindir=/sbin --with-staticdir=/sbin CLDFLAGS=-Wl,-O1

==> --with-clvmd=none is that normal?

> 
>> sys-fs/multipath-tools-0.4.7-r2 (0.4.8 doesn't work)

I have added a very "usefull" (for me) bug report for multipath-tools :
http://bugs.gentoo.org/show_bug.cgi?id=293615

multipath hangs IOs instead of marking a volume as unusable... If I unmap a volume, multipath try changing the path all the time. So a raid1 on 2 volumes is not useable.

Re: [gentoo-cluster] clvm/lvm2/device-mapper

[Ulrich David]

After looking in detail clvm is not what I need... clvm provide a locking system for protecting lvm configuration changes on lv... It doesn't prevent mounting the volumes on 2 hosts.

After searching for a while, I don't find any reponse for my question "What do I need in order to protect my shared storage (shared SAN lun with lvm volumes) during live migration?". My question is probably not right... Is it correct to say that :
"A migration moves the complete domU state including the current VFS state.  The old domU is paused and all remaining dirty blocks are copied over before the migrated domU is taken active and the paused domU gets killed."

If it is I don't need to protect my shared lvm volume.... If nobody try to mount it on the two dom0 ?

David



Le 18 nov. 2009 à 19:31, Ulrich David a écrit :

> I just need to rebound on my initial question. I now want to install lvm2 with "cluster aware" tools (clvm). I want to use a lock on my lvm volumes which could be mount by multiple hosts (but only one at a time).
> I unmask clvm in the profile file of amd64. But when I install lvm2 with clvm use flag I don't have a useable clvm :
> 
> iceage0 ~ # /etc/init.d/clvmd start
> * Caching service dependencies ...
> *  Can't find service 'cman' needed by 'clvmd';  continuing...                                                                                                                                  [ ok ]
> cman
> * ERROR:  Some services needed are missing.  Run
> *         './clvmd broken' for a list of those
> *         services.  clvmd was not started.
> iceage0 ~ # /etc/init.d/clvmd status
> * status:  stopped
> 
> If I unmask cman, the result is :
> 
> 
> iceage0 ~ # emerge -pv lvm2
> 
> These are the packages that would be merged, in order:
> 
> Calculating dependencies... done!
> 
> emerge: there are no ebuilds to satisfy "=sys-cluster/cman-2*".
> (dependency required by "sys-fs/lvm2-2.02.51-r2" [ebuild])
> (dependency required by "lvm2" [argument])
> 
> no cman-2 in portage...
> 
> I have noticed that with clvm flag, configuration option for lvm2 are :
> * econf: updating LVM2.2.02.51/autoconf/config.sub with /usr/share/gnuconfig/config.sub
> ./configure --prefix=/usr --build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu --mandir=/usr/share/man --infodir=/usr/share/info --datadir=/usr/share --sysconfdir=/etc --localstatedir=/var/lib --enable-readline --disable-selinux --enable-pkgconfig --libdir=/usr/lib64 --enable-dmeventd --enable-cmdlib --enable-applib --enable-fsadm --enable-static_link --with-mirrors=internal --with-snapshots=internal --with-lvm1=internal --with-cluster=internal --with-clvmd=none --with-pool=internal --sbindir=/sbin --with-staticdir=/sbin CLDFLAGS=-Wl,-O1
> 
> ==> --with-clvmd=none is that normal?

[gentoo-cluster] unsubscribe

[Guy Harrison]