Enterprise Volume Management System

On Thursday 30 May 2002 19:33, Mark Peloquin wrote:
> I think EVMS may be getting confused by finding LVM
> metadata on both RAID device as well as the raw
> disks.
>
> I need more info to debug this further. Could you change
> the kernel info logging level to a higher verbosity? Do
> this by entering the following command:
>
> evms_info_level 6       #this will increase the logging level
> evms_rediscover         #make the kernel try to rebuild its volumes
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D info_leve 6 =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D
May 31 04:01:03 boot-cd kernel: evms: evms_full_rediscover_prep: started.
May 31 04:01:03 boot-cd kernel: evms: evms_full_rediscover_prep: complete=
d.
May 31 04:01:03 boot-cd kernel: evms: dos_part: exporting segment=20
'ataraid/d01'.
May 31 04:01:03 boot-cd kernel: evms: dos_part: exporting segment 'hdg1'.
May 31 04:01:03 boot-cd kernel: evms: dos_part: exporting segment 'hde1'.
May 31 04:01:03 boot-cd kernel: evms: md core: autorun ...
May 31 04:01:03 boot-cd kernel: evms: md core: ... autorun DONE.
May 31 04:01:03 boot-cd kernel: evms: md core: EVMD MD:autostart_arrays()=
 EXIT=20
(exported_nodes=3D0)
May 31 04:01:03 boot-cd kernel: evms: lvm: Group raid1 is incomplete.
May 31 04:01:03 boot-cd kernel: evms: lvm:      Only 3 of 1 PVs found.
May 31 04:01:03 boot-cd kernel: evms: lvm:      Volumes in this group may=
 be=20
incomplete.
May 31 04:01:03 boot-cd kernel: evms: lvm: Exporting LVM volume lvm/raid1=
/root
May 31 04:01:03 boot-cd kernel: evms: lvm: Exporting LVM volume lvm/raid1=
/var
May 31 04:01:03 boot-cd kernel: evms: lvm: Exporting LVM volume lvm/raid1=
/tmp
May 31 04:01:03 boot-cd kernel: evms: lvm: Exporting LVM volume lvm/raid1=
/swap
May 31 04:01:03 boot-cd kernel: evms: md core: autorun ...
May 31 04:01:03 boot-cd kernel: evms: md core: ... autorun DONE.
May 31 04:01:03 boot-cd kernel: evms: md core: EVMD MD:autostart_arrays()=
 EXIT=20
(exported_nodes=3D0)
May 31 04:01:03 boot-cd kernel: evms: lvm: Group raid1 is incomplete.
May 31 04:01:03 boot-cd kernel: evms: lvm:      Only 3 of 1 PVs found.
May 31 04:01:03 boot-cd kernel: evms: lvm:      Volumes in this group may=
 be=20
incomplete.
May 31 04:01:03 boot-cd kernel: evms: md core: autorun ...
May 31 04:01:03 boot-cd kernel: evms: md core: ... autorun DONE.
May 31 04:01:03 boot-cd kernel: evms: md core: EVMD MD:autostart_arrays()=
 EXIT=20
(exported_nodes=3D0)
May 31 04:01:03 boot-cd kernel: evms: lvm: Group raid1 is incomplete.
May 31 04:01:03 boot-cd kernel: evms: lvm:      Only 3 of 1 PVs found.
May 31 04:01:03 boot-cd kernel: evms: lvm:      Volumes in this group may=
 be=20
incomplete.
May 31 04:01:03 boot-cd kernel: evms: deleting duplicate reference to=20
'lvm/raid1/root'.
May 31 04:01:03 boot-cd kernel: evms: deleting duplicate reference to=20
'lvm/raid1/var'.
May 31 04:01:03 boot-cd kernel: evms: deleting duplicate reference to=20
'lvm/raid1/tmp'.
May 31 04:01:03 boot-cd kernel: evms: deleting duplicate reference to=20
'lvm/raid1/swap'.
=3D=3D=3D=3D=3D=3D=3D=3D=3D /info_leve 6 =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

Hope that helps. More informations in attached file (log level 8).=20

Thanks

Thomas

The relevent portion of the log you sent me follows:

May 31 04:03:21 boot-cd kernel: evms: lvm: Checking Group raid1
May 31 04:03:21 boot-cd kernel: evms: lvm: Re-reading PE maps for node hde1
May 31 04:03:21 boot-cd kernel: evms: lvm: Re-reading PE maps for node hdg1
May 31 04:03:21 boot-cd kernel: evms: lvm: Re-reading PE maps for node
ataraid/d01
May 31 04:03:21 boot-cd kernel: evms: lvm: Group raid1 is incomplete.
May 31 04:03:21 boot-cd kernel: evms: lvm:      Only 3 of 1 PVs found.
May 31 04:03:21 boot-cd kernel: evms: lvm:      Volumes in this group may
be incomplete.

What this is saying is that the LVM plugin found 3 exact copies of
metadata.
One on the RAID device, and one of each of the mirrors. LVM only expected
one, so it complained.

For now, I'm afraid to say that, EVMS will not work correctly with ATARAID
because it sees the same metadata on both the produced RAID entity
and the consumed IDE entities, which will confuse most, if not all, plugins
that lay metadata within them. ATARAID attempts the hide the partitions
on the consumed entities by modifying some in memory data structures.
Unfortunately, since EVMS reads the partition tables directly, it still
sees
the partitions and produces, in your case, hde1 and hdg1.

I'm now attempting to better understand how ATARAID works to
determine the best approach for enabling EVMS to utilize these
devices.

Mark

On Thursday 30 May 2002 16:46PM, Thomas Lohmiuller wrote:
> > For now, I'm afraid to say that, EVMS will not work correctly with
ATARAID
> > because it sees the same metadata on both the produced RAID entity
> > and the consumed IDE entities, which will confuse most, if not all,
plugins
> > that lay metadata within them. ATARAID attempts the hide the partitions
> > on the consumed entities by modifying some in memory data structures.
> > Unfortunately, since EVMS reads the partition tables directly, it still
> > sees
> > the partitions and produces, in your case, hde1 and hdg1.
> >
> could it be a solution for me to comment following line out?
> BLK_DEV_INFO( IDE2_MAJOR, "hd", "e", TYPE_IDE, INDEX_ALPHA );

Good idea! Remember to remove the line for "hd","g" as well,
then only the RAID device will be seen and you should be up
and running.

This is custom solution that should work for YOU with your
current configuration. If your configuration should change,
you may be required to tweak ldev_mgr.c accordingly to keep
things working.

> Is there any chance that this temporary solution would be compatible with
> future versions of evms?

Until a generic ATARAID solution is available, you will have
to modify ldev_mgr.c each time you upgrade EVMS.

> I really like evms and hope it's possible to use it this way.

Thanks for the kind words. I'm hoping we can come up with
something soon to at least make EVMS properly recognize
these devices.

Mark