Computers won't boot to FOG due to DHCP issue (maybe)?

[loosus456]

Server

• FOG Version: 1.3.5
• OS: CentOS 7

Client

• Service Version:
• OS:

I’ll probably update FOG later, but I just can’t right now. I am using the 4.11.6 TomElliott (x64) kernel.

This is a known working FOG server; it’s been used in other areas.

I have never used FOG to deploy to/from on this model of machine, though: Dell Optiplex 3030 AIO.

I also have never used FOG in this exact room, for what it’s worth.

None of the PCs in this room will boot to FOG (unless I set a static IP address). Here are the steps I’ve taken:

1. Attempt to boot to FOG using legacy boot.
2. This appears and takes normal than usual but eventually goes through:
   
3. After that, it fails when trying to configure the interface:
   
4. I type “s” to go into the iPXE CLI and assign a static IP and also try pinging another computer; the pings are successful:
   
5. Then, I type exit to exit from the iPXE CLI, and I can then get into FOG as usual.
6. Just as a test, I try to do a full-host inventory. As it’s coming up, this errors appear a few times:
   
7. Nonetheless, it eventually does get to the inventory prompt. I fill in the usual inventory info, and it sends it to FOG. Everything is fine.

These computers have an existing Windows installation on them. If I boot the the computer into Windows, DHCP seems to be working just fine, renewing its IP lease very quickly.

Any ideas?

[george1421]

@loosus456 If you have not resolved this, then make sure your network team has set the building switch to use one of the fast spanning tree procools like (FastSTP, RSTP, or what ever your switch mfg calls it). This will allow the port to start forwarding data then detect if there is a loop back second. (pessimistic vs optimistic forwarding)

[loosus456]

@george1421 Thanks for the help.

It’s mostly a spanning tree issue.

There also appears to be some kind of weirdness with thie specific hardware. But it seems so far that updating the BIOS clears that up (after spanning tree is appropriately set).

[george1421]

@loosus456 Yes the switch the computer is plugged into. Insert an unmanaged switch between the target computer and current switch being used. The fog server end is ok so no change is needed.

[loosus456]

@george1421 It takes about 55 seconds from power on until a usable desktop appears.

When you say “building switch,” do you mean the switch that the computer is connected to?

[george1421]

@loosus456 said in Computers won't boot to FOG due to DHCP issue (maybe)?:

If I boot the the computer into Windows, DHCP seems to be working just fine, renewing its IP lease very quickly.

Not to poke MS Windows, but by the time MS Windows boots, FOS would have already started imaging a computer. iPXE takes about 3 seconds to boot, and FOS is about 8 seconds. During each transition the link light winks starting the 27seconds over again. How long does it take to reboot windows (not that fancy sleep mode), but a real load the OS from disk?

[george1421]

It sure sounds like a spanning tree issue.

@loosus456 I want you to test this idea by placing an unmanaged switch between the pxe boot computer and the building switch. That unmanaged switch will keep the building switch port from winking while the pxe booting client starts up.

Another way to test if its spanning tree, is in your step 4 instead of doing a ifopen issue a dhcp net0 (I think) but wait at least 30 seconds at the iPXE> prompt. Standard spanning tree takes 27 seconds from the time the link light goes on until it starts forwarding data. (the unmanaged switch test is easier if you have one).

[Wayne Workman]

@george1421 do you think this sounds like the many portfast & spanning-tree issues we’ve had in the past?