need an ipxe debugging guide

[anderbubble]

So after working for some time, ipxe has stopped working for our little test cluster and I can’t quite figure out the software changes that prevent it from working. The first stages of the ipxe boot seem to go okay, the dhcp server responds, it configures the net0 device, deices on the filename it wants ( http://(server-local-ip):9873/ipxe/(node-local-mac-address)” with the server local IP and node local MAC address corret, then goed into “connecting” but eventually times out. Trying to fetch the file directly by curl or wget from the localhost returns nothing. Not sure what I broke but dar, ipxe debugging is an obscure art!

See also https://medium.com/@panda1100/warewulf-v4-ipxe-node-provisioning-explained-189422b10486

[griznog]

This is the behavior I see from recent ipxe on some NICs, usually Mellanox ConnectX cards. The fix I've been using is to roll ipxe back to the commits in August 2023, there is some commit after that which causes this network timeout to start happening for some cards. It's on my to-do list to narrow this down to a specific commit and report it to ipxe upstream, but not near enough to the top of that to-do list.