iPXE stalls during HTTP ISO Boot when accessing an LACP connected server.

[MultimediaMan]

Pretty much as the title says:

In my home lab, I have an LACP-Capable switch (HP Procurve J9028B), which has a webserver which has a pair of interfaces configured for LACP. The corresponding switch ports are also configured for LACP.

When I SANBoot an ISO larger than about 200MB (which has an embedded ISOLINUX/mboot.c32 boot kernel) I get an odd slowdown, which finally grinds to a halt... a bit of patience revealed this to be a pause of about 90-300 seconds (several times, depending on the image size) during the boot. The boot never fails, but it pauses in the exact same spots each time. The stall occurs during the "append" sequence in the isolinux.cfg file.

This slowdown/stall does not occur on a non-LACP Interface, nor from a Virtual running on an LACP-connected Physical Host. Very Curious.

Where to begin with troubleshooting this beast? I've even gone so far as to break the LACP Team on the Webserver - the problem goes away when I do.

TIA,

Matt

[jpwise]

Not my area of expertise, but it does sound very similar to some issues I encountered a couple of weeks ago setting up an iPXE boot server in one of our other branches. After doing a tcpdump we traced it to the Cisco switch using STP (Spanning Tree Protocol) resulting in a ~60 second delay with bringing the port online after initial link.

Similarly to what you experienced it wasn't an issue from a VM running in bridge mode if the host machine was already connected to the port (and having sat through the delay already).

I'd speculate it's probably a similar issue with the switch checking for a loop to prevent flooding the network. Could give you a starting point to troubleshoot it further at any rate.

From memory if you plug a dumb switch into the port and then plug the iPXE client machine in 'behind' the dumb switch that also worked.

Jp.