Re: [SLUG] Booting from software RAID 1

From: craig@ctrust.com
Date: Wed Jun 16 2004 - 09:32:08 EDT


On 16 Jun 2004 at 8:46, Larry Brown wrote:

> Have you looked into creating a script that assigns a swap file to a
> file on the mirror after the machine has booted and remove it on
> shutdown? If that is possible, it would work around the problem. I
> would think you would take a pretty good performance hit though. I
> didn't know that a problem disk where the swap file is would cause a
> failure to boot. But it does make sense.
>
> You might want to remove the swap file from fstab and simply mount it
> with your own script after a successful boot. Have a shutdown sequence
> that places a tag file that the script looks for when mounting the
> swap. If the script fails to find the tag file (meaning it was an
> unexpected shutdown) it could simply not mount the swap and send an
> alert to your cell etc. If it does find the tag file it mounts the swap
> and removes the tag file. That way the server can still be running
> until you can arrive to take the necessary steps to verify the swap was
> not to blame etc.
>
> Just some thoughts...
>
> Larry
>
> On Tue, 2004-06-15 at 21:21, Ken Elliott wrote:
> > I usually set my Windows servers for RAID 1 (mirroring) on the boot disk.
> > But it seems Linux won't let me do this. Perhaps I'm doing it wrong, but
> > I've found no docs on this.
> >
> > Here's what seems to work:
> > Set /boot on it's own partition, and set RAID 1 on /, /usr and /home, etc.
> > Both disks have swap partitions, but if I crash either disk, the loss of
> > swap will kill the box. So, to get the real benefits, I must be able to
> > mirror Swap, or do without it (not good). If the loss of a disk crashes the
> > server, I might as well skip mirroring and setup a cron job to copy changed
> > files ever so often...
> >
> > I did find a site where a fellow put together a lengthy process involving
> > more drives and a lot of work. But it was written 4 years ago for kernel
> > 2.2. I'm on 2.6.
> >
> > Oh, yeah... The "hardware" RAID in the Intel 865 chipset is software
> > dependent - not true HW RAID.
> >
> > Thoughts?
> >
> > Ken Elliott
> >
> > =====================
> >
> > -----------------------------------------------------------------------
> > This list is provided as an unmoderated internet service by Networked
> > Knowledge Systems (NKS). Views and opinions expressed in messages
> > posted are those of the author and do not necessarily reflect the
> > official policy or position of NKS or any of its employees.
> >
>
> -----------------------------------------------------------------------
> This list is provided as an unmoderated internet service by Networked
> Knowledge Systems (NKS). Views and opinions expressed in messages
> posted are those of the author and do not necessarily reflect the
> official policy or position of NKS or any of its employees.

What kind of controller is on the intel boards? I have a number of
VIA controlled boxen in the field that are using high point controllers.
They don't seem to have any problem dealing with RAID. Linux
doesn't even know it is there (as it should be) The disks register
either as SCSI disks or an ATARAID disk (depending on kernel
version). You may consider a PCI IDE raid controller if these boxen
are very important. Adaptec makes a very nice one, I'm pretty sure it
is a 1200 or something like that. There may be a newer unit out
there that will deal with ATA133, but be careful to make sure it has
linux support.

Servers should NEVER have a software RAID for the very reason
you are having trouble.

-Craig
-----------------------------------------------------------------------
This list is provided as an unmoderated internet service by Networked
Knowledge Systems (NKS). Views and opinions expressed in messages
posted are those of the author and do not necessarily reflect the
official policy or position of NKS or any of its employees.



This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 17:45:09 EDT