[Beowulf] RE: A couple of interesting comments

Greg Keller Greg at keller.net
Wed Sep 24 07:37:12 PDT 2008


As a former installer/patsy at one of those nameless clumsy hardware  
vendors, I thought this *may* be useful for you:

> 1.  We specified "No OS" in the purchase so that we could install  
> CentOS
> as our base.  We got a set of systems with a stub OS, and an EULA for
> the diagnostics embedded on the disk.  After clicking thru the EULA,  
> it
> tells us we have no OS on the disk, but does not fail to PXE.

Vendors get into all kinds of legal issues when they don't sell M$  
licenses they are supposed to have EULA's for cover.  They used to  
clobber/corrupt partion tables etc. when fdisk'd ruthelessly but are  
largely benign now if you handle issue #2 gracefully and get the  
standard install/fdisk launched.  The vendor I worked for had a EULA  
free sku but only if you purchased RHEL WS licenses.  Blank disks  
Always had EULAs.

> 2.  BIOS had a couple of interesting defaults, including warn on
> keyboard error (Keyboard?  Not intentionally.  This is a compute node,
> and should never require a keyboard.  Ever.)  We also find the BIOS is
> set to boot from hard disk THEN PXE. But due to item 1, above, we  
> never
> can fail over to PXE unless we load up a keyboard and monitor, and hit
> F12 to drop to PXE.

A fast trick I've used:
a. Keyboard errors are usually just complaint messages on a console  
you'll never see.  I haven't seen them halt a system during boot in  
years so hopefully this is cosmetic.
b. to get around keyboarding through the BIOS config, a system with  
hot swap drives can be simple:
	Eject the drives
	Boot the system (with ethernet cabled to a powered on switch)
	Wait long enough for the system to post and decide there's no hard  
drive, which bumps the PXE up to first boot
	Power down, reinstall the drives - Now PXE is the primary boot device

This sounds more painful than it is.  I've deployed a few 500+ and a  
couple 1200+ node clusters and life was much faster and simpler this  
Note:  With many systems you can switch back to Disk as primary boot  
device by powering off the ethernet switch and going through the 2  
boot dance so the local drive becomes the primary boot device again.   
We normally setup PXE to tell the node to boot local on the occasions  
we need a stateful install.  YMMV

Once the nodes are booted you can script the "Config" utility from the  
vendor to tune the bios/bmc etc to your preferred favorite settings.   
I use Perceus so the local drives are always scratch+swap space and we  
partition them and format on the way up if they need it.

Hope this helps,

More information about the Beowulf mailing list