Trying to "Push" CentOS7 image on Lenovo P620
-
@hybride said in Trying to "Push" CentOS7 image on Lenovo P620:
iPXE 1.20.1+ (g4bd0)
Ok, now you are using the iPXE binaries we shipped with FOG 1.5.9. Definitely getting closer but not there yet.
Take a look at wireshark as George suggested and if that doesn’t get us to a solution you might still try compiling the latest iPXE source code and see if that binary works for your hardware.
-
@sebastian-roth Hi!
I’m struggling a bit with Wireshark as I can’t see or catch any packets on thoses ports (maybe it’s me who is not using wireshark correctly)
But I was able to get into à kinda shell (ipxe one) when booting and I can confirm that the computer is getting the proper IP (from the IP reservation pool). Here’s what I get:
Received DHCP answer on Interface net0
tftp://MY FOG IP SERVER/default.ipxe … Connection timed outCould it be this file that is older also?
-
@hybride said in Trying to "Push" CentOS7 image on Lenovo P620:
But I was able to get into à kinda shell (ipxe one) when booting and I can confirm that the computer is getting the proper IP (from the IP reservation pool). Here’s what I get:
Received DHCP answer on Interface net0
tftp://MY FOG IP SERVER/default.ipxe … Connection timed outWell that’s interessting. Seems like it is actually receiving an IP from the DHCP but then fails to download default.ipxe. The file is just a text based iPXE config file. You can simply edit it (
/tftpboot/default.ipxeon your FOG server). Don’t think there is an older version of it that could cause such an issue.From what you said other machines (e.g. legacy BIOS based ones) properly PXE boot to FOG, right? So the connection timeout we now see is not a general issue but only happens with the P620, right?
Wireshark would definitely be helpful to get a better understanding of what is happening on the network layer. Keep playing with it until you get the packets captured. On which machine do you run Wireshark? While some of the network communication is broadcasted over the whole subnet and can be captured anywhere - most traffic is unicast and needs to be captured at the communication endpoints (FOG server, easy - or client, harder) or in between (using an old network hub or monitoring port on a modern manageable switch).
The other option you still have is compiling iPXE binaries from the latest iPXE code - find the instructions in one of my earlier posts.
-
@sebastian-roth said in Trying to "Push" CentOS7 image on Lenovo P620:
Seems like it is actually receiving an IP from the DHCP but then fails to download default.ipxe.
I confirm that is what is happening, when I get into de iPXE shell and do “show ip” I do get the IP from my dhcp server.
my defautl.ipxe file contain this:
#!ipxe
cpuid --ext 29 && set arch x86_64 || set arch ${buildarch}
params
param mac0 ${net0/mac}
param arch ${arch}
param platform ${platform}
param product ${product}
param manufacturer ${product}
param ipxever ${version}
param filename ${filename}
param sysuuid ${uuid}
isset ${net1/mac} && param mac1 ${net1/mac} || goto bootme
isset ${net2/mac} && param mac2 ${net2/mac} || goto bootme
:bootme
chain http://MYFOGServerIP/fog/service/ipxe/boot.php##params
Indeed everything look fine (I guess)
From what you said other machines (e.g. legacy BIOS based ones) properly PXE boot to FOG, right? So the connection timeout we now see is not a general issue but only happens with the P620, right?
Yes, P620 (who does not have the option to change to “bios” mode. From there now I know that this FOG server most probably never worked on any kind of UEFI system.
Think I’m going to install Wireshark directly on the fogserver (CentOS server) If it can be done.
The other option you still have is compiling iPXE binaries from the latest iPXE code - find the instructions in one of my earlier posts.
I’ll have a look on this also
Can I do it on a side VM and then just copy all the “new” tftp folder on my Fog server?Thank you for your help! Very cool!
-
@hybride said in Trying to "Push" CentOS7 image on Lenovo P620:
Can I do it on a side VM and then just copy all the “new” tftp folder on my Fog server?
Sure! You actually don’t even need to copy all of the new compiled binaries. Just copy over the new
ipxe.efiand see if that makes a difference. If not also try the newsnponly.efijust to be sure. -
@sebastian-roth said in Trying to "Push" CentOS7 image on Lenovo P620:
n’t even need to copy all of the new compiled binaries. Just copy over the new ipxe.efi and see if that makes a difference. If not also try the new
Saddly I don’t have much luck with new compiled efi files
I’ll install Wireshark on my FOG server and see what I can track,
I’ll be back!
-
@sebastian-roth Here’s my wireshark capture file
Cheers!
-
@hybride said in Trying to "Push" CentOS7 image on Lenovo P620:
Here’s my wireshark capture file
Ok, that’s a good start. We see the TFTP transfer of
snponly.efifile. Though there is nothing else in the Wireshark dump. No DHCP packets nor any further TFTP request.Any chance you can get hold of an old network hub that you can connect between the Lenovo P620 and the network layer switch. Then connect another Windows notebook to that hub and capture the network traffic.
-
@sebastian-roth said in Trying to "Push" CentOS7 image on Lenovo P620:
o DHCP packets nor any further TFTP request
My fault I filtered my capture, here’s one with no filter in it.
Hope it does the job!
No filter this time
-
@Hybride In Wireshark use the display filter
eth.addr == e0:4f:43:e6:4a:3eto see all the packets sent from and received by the Lenovo P620 laptop. After the TFTP transfer (snponly.efi) we see packets of the second DHCP DORA done by iPXE (first one is done by the BIOS/UEFI firmware). And finally after that we see ARP resolution requests over and over.To me this looks like iPXE loaded on the L620 is able to actually send packets on the wire (as we see them arrive on the server side) but has problems to receive the answers. Otherwise it shouldn’t send the same ARP request so many times!
Still question is, why is it able to do the DHCP DORA (two packets sent and two packets received) but then fails to go ahead?!
Can you get to the iPXE shell again? Try sending a ping to the server:
ping x.x.x.xand let us know if this works. -
@sebastian-roth said in Trying to "Push" CentOS7 image on Lenovo P620:
(two packets sent and two packets received) but then fails to go ahead?!
I can’t ping anything, not even my gateway. I only get timeout.
-
@sebastian-roth Hello, just a minor correction: It’s a Lenovo Desktop P620 and the network card is a: Marvel AQtion 10Gbit Network Adapter
-
@Hybride Any chance you can get hold of an old network hub that you can connect between the Lenovo P620 and the network layer switch? Then connect another Windows laptop to that hub and capture the network traffic using Wireshark in that laptop.
-
@sebastian-roth Sadly I don’t have any at my office
-
@sebastian-roth I will check what I can do with some “port mirroring” maybe!
-
@sebastian-roth Well that was easier than I thought!
Here’s the file. But I’ve filtered the result a bit with this filter:
eth.addr == e0:4f:43:e6:4a:3eBecause there was a lot of stuff running on my laptop but just tell me if you prefer more detailled logs. But I would like to send the complete log to your email and not post it on the forum please.
-
@hybride said in Trying to "Push" CentOS7 image on Lenovo P620:
Here’s the file. But I’ve filtered the result a bit with this filter: eth.addr == e0:4f:43:e6:4a:3e
Can’t make sense of this with the filter enabled. Please check the speech bubble in the top right corner of the forums for a private message.
One thing I really wonder, are FOG server and client host in the same subnet or is there a router in between? What’s the subnet mask in your setup, 255.255.0.0?
As well what’s the MAC address of the client host? I thought it’d be e0:4f:43:e6:4a:3e but then I just figured we have ICMPv6 packets in the dump as well, send from this MAC. That doesn’t make sense to me.
-
@sebastian-roth Sorry for the delay!!! My HR decided I needed more jobs lol
So I was forced to push back my ipxe debugging. Between two users I’ve installed a Startech (20 box) 1Gb ethernet card on my Lenovo P620 and put snponly.efi (fog) file in the option 067 on my DHCP and now I can deploy.
I wish I knew how to put the AQtion card drivers in those Kernel, it’s always a Driver bug
Thank you very much for answering all my noob question tho !!
If you want me the sent you drivers files or something to update Kernel just let me know, I’ll be happy to send you whatever you need to update FOG
Best regards!
-
@Hybride Thanks for the update on this. Definitely a driver issue but as it’s in iPXE there is not much we can do. You might think about reporting to the iPXE developer mailing list if you are keen. But as you were able to get around this by using a different NIC it’s probably not worth the time and effort.
-
@sebastian-roth said in Trying to "Push" CentOS7 image on Lenovo P620:
if you are keen. But as you were able to get around this by using a different NIC it’s probably not worth the time and effort.
I’ve contacted Lenovo about this and here’s their answer:
"We are currently working with Marvell and the developers of iPXE to update the iPXE source code to include the Marvell NIC driver for the P620. Unfortunately, this hasn’t been a quick and easy fix and we are currently waiting on the iPXE developers to pick up the updated patch code provided by Marvell in their official iPXE release. In the meantime, I can provide you with a test iPXE image that should work for you on the P620 platform. "
I haven’t got the time to test is test iPXE but I’ll stick to the 20$ network card add
I rarely install more than 5 computers at a time so I’ll be fine this way.Thanks again Sebastian for your help,
