Re: RAID5 (mdadm) array hosed after grow operation (there are two of us)
On Thu, 30 Apr 2009, Boyd Stephen Smith Jr. wrote:
> He who codes, decides. Either put forth the effort to
> design/write/review/test/apply the patch or don't be surprised if your
> preferences are not highly weighted in the resulting code.
Will lvm upstream take something that makes lvm align pv data areas to RAID
stripe boundaries and is actually less arcane that hit and miss games with
pvcreate --metadatasize?
I have a lot of stuff on my plate right now (including updating some Debian
packages of mine that are getting cowebs and dealing with the entire mess
that the new kde did to my main workstation) and I can't waste effort on
something that wouldn't be accepted outright... even then, it will take a
while for me to actually get my hands dirty with lvm code.
> That said, I think everyone participating in the thread so far has agreed that
> we don't want separately-maintained detection code in LVM that could get out
> of date, but for LVM to use the existing detection code when it is available
> on the system.
Yes, please. mdadm is just as critical to system startup as lvm, there
isn't a problem into having dependencies between the two.
> >This kind of stuff really should not be done halfway, it can suprise someone
> >into a dataloss scenario.
>
> Right now, LVM will stomp all over devices with either 0.90 or 1.0
> superblocks. Even if it can only detect one or the other and not both, it
> will cause less data loss than now.
There is an old adage in the security circles about the dangers of false
sense of security, and it applies here. For me to agree with you, the
feature of not corrupting "inactive" md components would have to be
undocumented, so that nobody would trust it to exist and thus nobody would
be bitten by the fact that it is implemented partially.
--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh
Reply to: