Ronald G Minnich rminnich@lanl.gov writes:
This is with fallback and normal LinuxBIOS images on E7500 mainboards. They're loaded with the identical etherboot image, 5.1.2rc4.eb5.
On fallback, etherboot goes right to the IDE disk after finding no good ethernet link:
[snip]
On normal, etherboot just loops forever until I tell it to go to disk:
[snip]
It is a bug. And it is fixed in the etherboot source tree. Ron for your cluster I just need to make a new build of LinuxBIOS and put the new etherboot in there. The boot order code was just broken, and I don't know how it lasted as long as it did.
Ron can you check to see what your boot order is. As I recall it only happened with boot_first=Network boot_second=Network boot_third=Network
In which case there is a work around available. I need to track this down before we start building up Pink.
I'm guessing I'm not the first person who will hit this, hence this message to various lists.
The code was fixed a little while ago, when I completed the e1000 support. As the bug also it prevented fail over from one nic to another.
Eric