Menu
▾
▴
Re: [Devil-Linux-discuss] DL 1.2.10 with hard disk going slow...
|
From: Dom <dl...@ed...> - 2007-09-05 12:01:19
|
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Heiko Zuerker wrote:
<blockquote
cite="mid:542...@ww..."
type="cite">
<pre wrap="">On Mon, September 3, 2007 13:35, Dom wrote:
</pre>
<blockquote type="cite">
<pre wrap="">My DL-based samba server has worked like a dream for several years, now
it is on a serious go slow. Also I seem to have a lost an LVM logical
volume /var/data (filesystem ext2) - at boot there are messages something
like this: ---------------------------
fsck.ext2: Attempt to read block from filesystem resulted in short read
while trying to open /.../devil-linux/data Could this be a zero-length
partition? Mounting /var/data mount: you must specify the filesystem type
[FAILED]
---------------------------
But AFAIK I have never used /var/data so this shouldn't matter. The
other LVs (/home, /var/log, /opt, swap) seem to load OK but there are
usually transaction replays on /home (which is the important one with the
data) and /var/log.
And the samba server is running very slow. It has always been fast
before though I noticed a couple of months ago that it was occasionally
very slow accessing one sub-folder which had some very large files in it.
I have a very simple setup and haven't made any significant changes
recently. What is happening? My current theory is that the hard disk is
failing. I plan to (try to) migrate my data to a new RAID 1 two-disk array
and hope this gives me back the performance and gives better protection in
the event of failure.
But I am far out of my depth here. If anyone has any suggestions I would
be grateful to hear them!
</pre>
</blockquote>
<pre wrap=""><!---->
Your theory is most likely right, I would also point to the hard disk.
Make sure you got a good backup of everything, the disk may just stop
working completely soon or you may experience data loss.
Can you see anything in the syslog or on tty10 (alt + F10) ?
</pre>
</blockquote>
Thanks for your reply Heiko, and for confirming the diagnosis. You were
right about messages appearing on the local display - a series of
messages like this appeared on at least one occasion:<br>
end_request: I/O error, dev 03:00 (hda), sector 0<br>
<br>
So I:<br>
- took delivery of 2 new IDE PATA 160GB hard disks<br>
- configured them as software RAID1 following the DL instructions
(chapter 1 part 6), created a devil-linuz VG with the same LVs as the
existing disk [hardest bit for me was discovering the full name of the
RAID device to create the VG (<b>'</b>vgcreate -s 8M devil-linuz
/dev/md/0', because LVM1's vgcreate cannot use the /dev/md0 (or
/dev/hda etc)], checked this was all working okay. During this time the
old disk was still being used.<br>
- renamed the existing VG as devil-linux-old and renamed the new VG to
devil-linux (old disk no longer in use from this point of course as it
disappears from samba), rebooted<br>
- new devil-linux LVs are all auto-mounted, I manually mounted the old
home LV and then copied all files over to the new home LV<br>
<br>
Several hours later all seems to be working perfectly. I still don't
know quite what the original problem was, but it seems sorted and I
have more protection now (because of RAID1) against future hard disk
failure.<br>
<br>
I made one mistake (I now realise). I established the RAID1 device
directly off the drives not off the primary partitions. I typed:<br>
<span><b class="command"> mdadm --create /dev/md0 --chunk=64
--level=raid0 --raid-devices=2 /dev/hdb /dev/hdd<br>
</b><span class="command">instead of:<br>
</span></span><span><b class="command"> mdadm --create /dev/md0
--chunk=64 --level=raid0 --raid-devices=2 /dev/hdb1 /dev/hdd1<br>
</b><span class="command"><br>
I had created the partitions before doing this - I just got the typing
in the mdadm wrong (hdb instead of hdb1, hdd instead of hdd1). cat
/proc/mdstat reports:<br>
<br>
md0 : active raid1 ide/host0/bus0/target1/lun0/disc[0]
ide/host0/bus1/target1/lun0/disc[1]<br>
<br>
It all works fine but 'fdisk -l' reports that there are no partitions
on either of these drives (hdb or hdd). It seems to me that mdadm has
destroyed the partitions I created and is using the drives in some sort
of 'raw' unpartitioned state. Is this likely to cause me big problems
in the future? I'm uneasy that in the future someone (probably me)
might accidentally reformat one of these unpartitioned drives, not
realising that they are part of the RAID array. How would I undo this
(i.e. change the RAID device so it lies over partitions and not over
raw drives)? I presume that if I create partitions on the drives now I
will just destroy the information. But perhaps I can do it for one
drive/partition at a time and mdadm will rebuild the info?<br>
<br>
I've updated my instructions page which is really for my own benefit
but might help others here: <a class="moz-txt-link-freetext" href="http://www.edendevelopments.co.uk/dlsetup">http://www.edendevelopments.co.uk/dlsetup</a><br>
<br>
Thanks again!<br>
<br>
Dom<br>
</span></span>
</body>
</html>
|
