Linux Debian 5 on PE R610 : random freeze at boot

ROUSSEL Kévin Kevin.Roussel at loria.fr
Thu Sep 17 05:15:05 CDT 2009


Le mercredi 16 septembre 2009 à 10:10 -0600, Stephen Dowdy a écrit :
> Kévin
> 
> We're running Lenny on several R610/R710 systems here.  We got them
> before BIOS 1.1.4 (was 1.0.4) and had lots of grief.  several systems
> were installed 32-bit because, while the e820 BIOS memory probe stuff
> was broken (in BIOS 1.0.4), we could at least boot the systems in
> 32-bit, but not 64-bit.  (albeit with only 1GB of the RAM being
> configured :-( )
> 
> Now that 1.1.4 is out we still have some less major issues and have
> a few installed with amd64.
> 
> We disable the hpet timer because it locks up the machine sometimes
> for extended periods.  (you'll see lines like:
>     CE: hpet increasing min_delta_ns to 15000 nsec
> in 'dmesg' output)
> 
> Here's a script to disable the HPET timer:
> ---------------------
> grub_menu=/boot/grub/menu.lst
> sed -i-`date +%Y%m%d%H%M`.bak -e '/^#[[:space:]]*.*hpet=/,$b;s/^\(#[[:space:]]*kopt=.*$\)/\1 hpet=disable/' ${grub_menu}
> /usr/sbin/update-grub
> ---------------------
> 
> Here's another sed script to enable the rootdelay option programmatically, too:
> ---------------------
> grub_menu=/boot/grub/menu.lst
> sed -i-`date +%Y%m%d%H%M`.bak -e '/^#[[:space:]]*kopt=.*rootdelay=/,$b;s/^\(#[[:space:]]*kopt=.*$\)/\1 rootdelay=10/' ${grub_menu}
> /usr/sbin/update-grub
> ---------------------
> (these scripts only check if you have specified an hpet or rootdelay option,
> and assume if you have you know what you're doing, otherwise, they change
> it to use hpet=disable or rootdelay=10)
> 
> and i'm still not convinced that BIOS 1.2.6 kit for R610 is good.
> It appears to have a PowerEdge R710 BIOS payload in it.  That may
> be all well/good (the image may be the same), but i'm certainly
> not comforted by the filename, nor the fact that only "PowerEdge R710"
> as a string exists in the file, but not R610.
> 
> I have a support call into Dell over this,
> and after 2 weeks, still no answer (though frontline support has
> been quick to respond, as well as my program manager)
> 
> ---------------------
> pick:R610# wget http://ftp.us.dell.com/bios/PER610_BIOS_LX_1.2.6.BIN
> pick:R610# sh PER610_BIOS_LX_1.2.6.BIN --extract FOO
> pick:R610# ls -l FOO/payload
> total 4104
> -rwxr-xr-t 1 root root 4194396 2009-07-17 11:57 r710-010206.hdr*
> 
> Say, wha!?!  'r710*', instead of R610 hdr.
> 
> pick:payload# strings - * | grep 610
> pick:payload# strings - * | grep 710
> Dell System PowerEdge R710
> ---------------------
> 
> So, i'm leery to apply/attempt this BIOS update myself until i
> hear affirmatively that Dell believes this is the correct setup.
> 
> From your last message, it seems to indicate that the
> system is up, but "crashing", (so presumably rootdelay=
> fixed your random boot failures?)
> 
> Anyway, hope this helps a bit.
> 
> --stephen


** Correction **

The 'rootdelay=9' boot option for the Linux kernel actually *works* for
two of the servers I manage. I have been able -- by programming cron to
reboot them every 15 minutes -- to make them reboot 60+ times without a
glitch.

The first machine, on which I tried this configuration trick without
success, is in fact suffering from an hardware problem (weakness of the
SAS backplane connection with motherboard).

It seems then that my problem is solved. Thanks to everyone!

----

For your interrogation about the BIOS update, I was indeed surprised by
the r710xxxx name for the HDR file, but it appears that the R610 and
R710 are part of a server line collectively known as Dell 11G servers,
and thus share the same BIOS.

I have flashed this HDR file to the BIOS of my three servers (using the
"dellBiosUpdat"' command of the Debian-stable "libsmbios-bin" package),
and I haven't encountered any problem since. My servers now announce me
they are using BIOS version 1.2.6, while they still know they are R610
servers (they haven't started believing they were another model, nor do
they suffer from another form of schizophrenia).

So I think this BIOS update is safe, even though Dell could indeed be
more informative about the update process...

Thank you,


    K. Roussel
    INRIA-Lorraine




More information about the Linux-PowerEdge mailing list