Just a heads up to anyone who might be running to some PXE boot issue
within RHEV/oVIRT environments when using a xinit.d tftp server running on
Centos/RHEL. For some reason the hypervisors are causing an issue with
TFTP and the negotiation of block size thus causing a timeout when
retrieving the pxelinux kernel. In a TCP dump you'll see the file
requested from the client ack'd by the server and the first block of the
file sent from the server to the client but its never ack'd and the server
continues to attempt to send the first block until it times out.
The fix for this is to add '-r blksize' to your command line arguments to
force a block size of 512. While there are more packets sent its still
very fast and successfully pulls the file.
Hopefully this saves someone from banging their head against the wall/desk
like I did