[prev in list] [next in list] [prev in thread] [next in thread]
List: linux-raid
Subject: Re: Partition table disappearing
From: Linas Vepstas <linas () fc ! net>
Date: 1997-09-29 18:33:21
[Download RAW message or body]
Ingo Molnar wrote:
> On Sat, 27 Sep 1997, John Lellis wrote:
>
> [...]
>
> > suggesting that fdisk overlaps the /dev/sdn1 and /dev/sdn2 partitions.
>
> [...]
>
> > "Begin" is the last sector on cylinder 1024 and "Start" is the first sector
> > on cylinder 1025.
> >
> > But why should this cause the partition table to disappear under RAID5? :-(
>
> because the RAID superblock (storing various RAID related on-disk
> information) is stored at the end of the partition. This is necessary to
> be able to mount RAID-1 filesystems under non-RAID kernels. Now, the
> problem is present under 'normal' ext2fs too, but in a much more subtle
> way: random disappearance of the partition table. (ext2fs allocates in a
> forward way, it might be that you dont see the problem for a _long_ time).
OK, please clarify. This sounds vaguely alarming. A couple of times a year,
my disk fills up due to some runaway script. Linux is pleasently robust with
this error: it doesn't die. I just kill the runaway script, delete and/or trim
the
offending file(s), and everything is back to normal.
Can I assume that this will still be the case with /dev/md0 mounted?
If so,then is this statement accurate:
if one half of a mirror is mounted (instead of /dev/md0), and the disk fills up,
then the raid superblock will be trashed?
>
>
> what does fdisk's 'x' say for such a parition table? It should complain
> ...
>
> -- mingo
--
Linas Vepstas -- linas@fc.net.no.spam.please -- http://linas.org/
[Attachment #3 (text/html)]
<HTML>
Ingo Molnar wrote:
<BLOCKQUOTE TYPE=CITE>On Sat, 27 Sep 1997, John Lellis wrote:
<P>[...]
<P>> suggesting that fdisk overlaps the /dev/sdn1 and /dev/sdn2 partitions.
<P>[...]
<P>> "Begin" is the last sector on cylinder 1024 and "Start" is the first
sector
<BR>> on cylinder 1025.
<BR>>
<BR>> But why should this cause the partition table to disappear under
RAID5? :-(
<P>because the RAID superblock (storing various RAID related on-disk
<BR>information) is stored at the end of the partition. This is necessary
to
<BR>be able to mount RAID-1 filesystems under non-RAID kernels. Now, the
<BR>problem is present under 'normal' ext2fs too, but in a much more subtle
<BR>way: random disappearance of the partition table. (ext2fs allocates
in a
<BR>forward way, it might be that you dont see the problem for a _long_
time).</BLOCKQUOTE>
OK, please clarify. This sounds vaguely alarming. A couple of times
a year,
<BR>my disk fills up due to some runaway script. Linux is pleasently
robust with
<BR>this error: it doesn't die. I just kill the runaway script, delete
and/or trim the
<BR>offending file(s), and everything is back to normal.
<P>Can I assume that this will still be the case with /dev/md0 mounted?
<BR>If so,then is this statement accurate:
<P>if one half of a mirror is mounted (instead of /dev/md0), and the disk
fills up,
<BR>then the raid superblock will be trashed?
<BR>
<BR>
<BLOCKQUOTE TYPE=CITE>
<P>what does fdisk's 'x' say for such a parition table? It should complain
<BR>...
<P>-- mingo</BLOCKQUOTE>
<PRE>--
Linas Vepstas -- linas@fc.net.no.spam.please -- <A \
HREF="http://linas.org/">http://linas.org/</A></PRE> </HTML>
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic