I'm running Windows Deployment Services on Windows Server 2008 R2 on top of an ESX 4.0 box. This is the only function of this VM instance, although it had previously functioned as an AD Domain Controller. My DHCP server is running on our primary Domain Controller, which is also Server 2008 R2, but running on metal. Everything was working perfectly until we recently had our backup generator fail during a power outage, causing all of our servers and networking equipment to lose power for a period of time. When we brought all of our equipment back up, everything was working as expected except for WDS.
Our network is split up into several different vlans. Now, depending on which vlan the client computer is on, it's behaving differently when attempting to PXE boot into WDS. Our servers are located on the 10.55.x.x vlan, which, due to the nature of it, has no DHCP server active in it. The first computer we plugged in happened to be in the 10.99.x.x vlan, which is supposed to be reserved for network management devices (i.e. switches), but we've been using it occasionally otherwise. That computer gave us PXE-E11 ARP Timeout errors. When we moved to a different computer on the 10.19.x.x vlan (for general purpose use), it finally gets an IP from DHCP, but it presents us with a very stumping PXE-E32 TFTP Open Timeout error. Before the power outage, it didn't matter which vlan a device was on; it would PXE boot and image just fine.
I've made no changes to anything server-side. Everything is configured exactly the same way it was on my WDS and DHCP servers as before the power outage. I've tried several different computers, including different models. All of this, combined with the quirky behavior depending on the vlan, makes me think something went wrong in one or more of our switches, probably because of the power outage. Unfortunately, I'm no network guy, and I know very little about how to configure our switches properly.
Is this an issue with switches, etc? If so, how can I fix it? Is there some magical option I'm not aware of? Does anybody out there have any hunches? I've pretty much exhausted my ideas.
- Our main switch is an HP Procurve 5406.
- We also have 3x HP Procurve 4208 switches.
- The ESX Server is an HP ProLiant DL380 G6.
- The WDS VM is currently using the VMXNET3 network adaptor, but we've also tried the E1000 adaptor.
Both issues are bugs in the PXE bootloader on HP servers (I've seen it on the ML150 and DL360 myself) when it's got to deal with a non-1500 MTU (i.e. there's a trunk somewhere). Here's how I've fixed it: