linux-lvm.redhat.com archive mirror
 help / color / mirror / Atom feed
* [linux-lvm] Reliability
@ 2000-06-07 17:07 Jan Niehusmann
  2000-06-07 17:43 ` Heinz J. Mauelshagen
  0 siblings, 1 reply; 4+ messages in thread
From: Jan Niehusmann @ 2000-06-07 17:07 UTC (permalink / raw)
  To: linux-lvm

What happens when one of several drives in a vg fails? Is the whole vg
lost or only lv's that are on that pv? In the latter case, what about
lv's that are only partly located on that pv?

I once observed that with a corrupted partition table, causing one pv
to be lost, the whole vg didn't get detected. As I was able to reconstruct
the partition table, I didn't try very hard to get the vg running 
without that drive. 

Jan

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [linux-lvm] Reliability
  2000-06-07 17:07 [linux-lvm] Reliability Jan Niehusmann
@ 2000-06-07 17:43 ` Heinz J. Mauelshagen
  2020-11-27 16:17   ` Michael Marxmeier
  0 siblings, 1 reply; 4+ messages in thread
From: Heinz J. Mauelshagen @ 2000-06-07 17:43 UTC (permalink / raw)
  To: Jan Niehusmann; +Cc: linux-lvm


Quorum support is a solution to deal with the activation of a VG without having
all PVs available which belong to it.

This actually is usefull in case you have associated all LEs of specific LVs
to available PVs and want to access those.
In real life scenarios this can't be garanteed though which leads to partitally
destroyed (unusable) LV contents and in consequence to filesystem errors.

Therefore qourum is IMO not really usefull.

The tradeoff of the existing solution nevertheless is that reconstruction of
the partition table and the VGDA is necessary on a replacement disk (subsystem)
which takes some time :-{(
After this we almost have the same scenario as with quorum.
The difference is that no I/O errors are returned in case of accessing
bad LEs.

Heinz

On Wed, Jun 07, 2000 at 07:07:57PM +0200, Jan Niehusmann wrote:
> What happens when one of several drives in a vg fails? Is the whole vg
> lost or only lv's that are on that pv? In the latter case, what about
> lv's that are only partly located on that pv?
> 
> I once observed that with a corrupted partition table, causing one pv
> to be lost, the whole vg didn't get detected. As I was able to reconstruct
> the partition table, I didn't try very hard to get the vg running 
> without that drive. 
> 
> Jan

-- 

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                Sistina Software Inc.
Senior Consultant/Developer                      Bartningstr. 12
                                                 64289 Darmstadt
                                                 Germany
Mauelshgen@Sistina.com                           +49 6151 710386
                                                      FAX 710396
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [linux-lvm] Reliability
  2020-11-27 16:17   ` Michael Marxmeier
@ 2000-06-08 10:11     ` Jan Niehusmann
  0 siblings, 0 replies; 4+ messages in thread
From: Jan Niehusmann @ 2000-06-08 10:11 UTC (permalink / raw)
  To: Michael Marxmeier; +Cc: Mauelshagen, linux-lvm

On Thu, Jun 08, 2000 at 02:31:19AM +0000, Michael Marxmeier wrote:
> As far as i can see Quorum support looks like the saner option:
> If you loose a disk and only restore the partition table and VGDA
> the replacement disk still mostly contains garbage (eg. not a fs).

That, and it may be usefull to be able to say: I know something is
wrong, but I need some files _now_, so please give me everything 
you have and don't wait until I mounted and initialized a new drive.

Perhaps damaged VGs should be read only, at least by default, to prevent
further damage. 

I was really worried when LVM didn't start with a corrupted partition
table: The PV on that disk was actually empty (perhaps a LV on it,
but no data) as I only added it to try out ext2 resizing without risking
the other partitons.

Jan

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [linux-lvm] Reliability
  2000-06-07 17:43 ` Heinz J. Mauelshagen
@ 2020-11-27 16:17   ` Michael Marxmeier
  2000-06-08 10:11     ` Jan Niehusmann
  0 siblings, 1 reply; 4+ messages in thread
From: Michael Marxmeier @ 2020-11-27 16:17 UTC (permalink / raw)
  To: Mauelshagen; +Cc: jan, linux-lvm

> Quorum support is a solution to deal with the activation of a VG without having
> all PVs available which belong to it.
> 
> This actually is usefull in case you have associated all LEs of specific LVs
> to available PVs and want to access those.
> In real life scenarios this can't be garanteed though which leads to partitally
> destroyed (unusable) LV contents and in consequence to filesystem errors.
> 
> Therefore qourum is IMO not really usefull.
> 
> The tradeoff of the existing solution nevertheless is that reconstruction of
> the partition table and the VGDA is necessary on a replacement disk (subsystem)
> which takes some time :-{(
> After this we almost have the same scenario as with quorum.
> The difference is that no I/O errors are returned in case of accessing
> bad LEs.

As far as i can see Quorum support looks like the saner option:
If you loose a disk and only restore the partition table and VGDA
the replacement disk still mostly contains garbage (eg. not a fs).
So we could as well activate the VG and leave LVs with extents
on offline PVs disabled?

Any reason why this might not work?

Michael

--
Michael Marxmeier           Marxmeier Software AG
E-Mail: mike@msede.com      Besenbruchstrasse 9
Phone : +49 202 2431440     42285 Wuppertal, Germany
Fax   : +49 202 2431420     http://www.msede.com/

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2020-11-27 16:17 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2000-06-07 17:07 [linux-lvm] Reliability Jan Niehusmann
2000-06-07 17:43 ` Heinz J. Mauelshagen
2020-11-27 16:17   ` Michael Marxmeier
2000-06-08 10:11     ` Jan Niehusmann

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).