Skip to content. | Skip to navigation

Personal tools


You are here: Home / Wiki / Kb200


Emulab FAQ: Testbed Operations: How do I deal with a "PXE Failed" message?

Emulab FAQ: Testbed Operations: How do I deal with a "PXE Failed" message?

> PXE Failed: _

Okay, this will probably be a ticky one to track down - There are numerous reasons this can happen.

First thing to do is look for the PCs MAC in the dhcpd log - /usr/testbed/log . Watch that file while the node is supposedly trying to PXE boot, and see if any requests get to the dhcp server.

If it gets a request, hopefully there will be some helpful message.

IIf not, I suspect a problem with your control net switch. Check the switch to see if it thinks it has carrier on that port, and check the speed/duplex settings to make sure they look right. If they're set to autonegotiate, try setting them to 100/full, or if they're hardwired, try setting them to auto.

Our Cisco switches here have a feature that when a port first comes up, for some period of time, they do not forward traffic for it - they are watching for loops in the topology. Of course, this means they drop PXE/dhcp packets, since those get sent first. We have to disable this feature on our switches, which is called "spantree portfast" (enabling portfast disables the feature.) I don't know what kinds of control net switches you have, so I don't know if this is an issue on them, or how you would deal with it if it is.

Finally, check the PXE configuration in the BIOS and on the card, and see if there's anything you can do to lengthen the timeout, etc.

Oh, and I guess you should make sure that the interface that is PXEing is the one plugged into the control net switch. :)