SCIENTIFIC-LINUX-USERS Archives

May 2005

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Connie Sieh <[log in to unmask]>
Reply To:
Connie Sieh <[log in to unmask]>
Date:
Mon, 9 May 2005 13:18:19 -0500
Content-Type:
TEXT/PLAIN
Parts/Attachments:
TEXT/PLAIN (89 lines)
Jon,

Since there were not many SATA raid devices availalbe when RHEL3 first 
was created I suspect that the code is a bit "young" in this area.  I 
have seen similar issues.  Granted RHEL Update x could have fixed this 
problem but Redhat has decided to not fix it.

-Connie
On Mon, 9 May 2005, Jon Peatfield 
wrote:

> This isn't (I'm sure) SL30x specific, but I just thought that someone else 
> might hit the same (mis-feature) and it might save a little time.
> 
> We just got a new "server" which originally came with an internal IDE disk
> (just for booting since I'm paranoid about booting from plug-in-cards) and 
> an aacraid controller.
> 
> Installing SL304 on this worked much as expected but it turns out that the
> PATA/IDE controller on this board doesn't support DMA (with the SL3
> kernels anyway).  So I added in a SATA boot disk (which is what the Vendor 
> originally was going to include).
> 
> Off went the install and I was initially surprised that the kernel found 
> the aacraid before the SATA disk but _I_ don't care about disk names...
> 
> However, it seems that anaconda guesses the bios disk order (as needed for 
> grub installation/config), based on:
> 
>   IDE disks first
>   SCS disks in the order they are _detected_
> 
> since the aacraid module is loaded before ata_piix (alphabetical?) it ends
> up guessing wrong for this box.
> 
> Ok, this is an unusual box since few machines will have all of PATA, SATA
> and aacraid devices.
> 
> However, testing on a "standard" (modern) desktop pc showed that iff it
> has both a SATA and PATA disk then again the order detected is wrong, ie
> it puts grub onto hda referring to hda as (hd0) and sda as (hd1) in the
> grub.conf.  The default bios boot order is of course sata first then pata 
> etc.
> 
> Now as in all things of this kind one can adjust the order that anaconda
> uses, e.g. for my server with aacraid I can just specify the disk order
> as:
> 
>    sdb, hda, sda
> 
> and all is well again, e.g. in a kickstart file:
> 
>   bootloader --driveorder=sdb,hda,sda
> 
> but I *think* that maybe SATA disks *should* default to being earlier than 
> PATA (most of the time).
> 
> Of course this won't always be right either 'cos people can change the
> bios settings, but I'd suspect that picking the defaults might be more
> useful than the current guessing.
> 
> I see that during the %pre section there is a file /tmp/scsidisks which 
> contains a list of disks and the driver they use, e.g.
> 
> sda  aacraid
> sdb  ata_piix
> 
> or (without aacraid) just
> 
> sda  ata_piix
> 
> etc.
> 
> Clearly I could have %pre create/modify a "--driveorder=" to put ata_*
> connected drives first, but:
> 
>   I may be missing something which will make this a really bad idea
> 
>   Someone else may already have a script to do this
> 
>   It may be that anaconda/kickstart has an option I missed
> 
> Any comments/suggestions?
> 
> Googling suggests that the code in SL4 will be quite different in this 
> area so it will only be a short-term fix anyway...
> 
> 

ATOM RSS1 RSS2