PXE on Dell M6600

[dhclose]

I'm trying to install F17 using PXE on a Dell M6600 laptop. We've been doing PXE installs on other Dell machines for quite a while, using F13 and F14 successfully. And the F17 installation on the M6600 seems to work. No errors are reported during the installation and no errors appear in the installation logs.

However, when we attempt to boot the newly installed system, the process gets almost all the way done, then utterly fails. Booting with the default options nearly fills the graphic balloon, then suddenly drops to these messages.

  Cannot open font file True
  Welcome to emergency mode. Use "systemctl default" or ^D to enter default mode

Neither choice actually works. No prompt is presented and no keyboard entry is accepted. (Keyboard interrupts are working as the NumLock light can be toggled properly.) The only exit is a power cycle.

Booting using grub to delete the rhgb and quiet options gets a number of errors.

  Failed to start Software RAID Monitor Takeover.
  Failed to start Configure read-only root support.
  Failed to start Load legacy module configuration.
  Failed to activate swap /dev/mapper/vg_laptop-lv_swap.
  Failed to start Initialize storage subsystems (RAID, LVM, etc.).
  Failed to start Initialize storage subsystems
  Failed to start Monitoring of LVM2 mirrors, snapshots, etc. using dmeventd or progress polling.
  Timed out waiting for device dev-mapper-vg_laptop\x2dlv_home.device

Followed by a huge number of "Job xxx failed with result 'dependency'." and "Dependency failed for xxx". Then it drops to the unresponsive emergency mode.

Booting in recovery mode gets most of the same errors, then also drops to the unresponsive emergency mode.

Booting the machine from the F17 netinst/recovery image on a USB stick is successful. Examination of the files shows an empty /var/log/messages and no hint of an error in any other log file. I have saved copies of the files in /var/log/anaconda.

After chroot, "rpm -V" shows a huge number (4591!) of missing files. Among the missing files are most of the /lib/modules and /lib/firmware directories, many of the /etc/systemd/system files, and many binaries in /usr/bin. (A complete list is available if needed.) It is surprising to me that booting gets as far as it does.

The problem is absolutely repeatable and identical on several different machines. If it is a hardware issue, then it must be a general characteristic of the M6600.

I'm stuck on what to do next and suggestions are appreciated.

[george_toolan]

Cannot open font file True

This should be a FAQ. See http://forums.fedoraforum.org/showthread.php?t=281278 and similar posts.

[dhclose]

Thank you, George. That eliminates one message. Unfortunately, it doesn't help with the basic problem, all the missing files. How is it possible for a PXE installation to complete without error and then to find 4500 files missing on first boot?

[RupertPupkin]

Hmm, that's weird. I did a PXE install of F17 on an old HP Pavillion zt1000 laptop with no problem, so I don't think it's a general F17 issue. Obviously the installation on your machine failed somewhere, despite the lack of indications in the install log. Was this an automated, unattended PXE install, or was it interactive? Because I would think the problem would be apparent during an interactive install.

[dhclose]

I don't think it's a general F17 issue.

I don't think so, either. When installed from DVD on the same machine, F17 is flawless.

Was this an automated, unattended PXE install, or was it interactive? Because I would think the problem would be apparent during an interactive install.

The whole point of PXE is to be automatic. We have several such machines.

The point of this thread and the reason for my question is to discover what else I can use to help isolate the issue and make things work. Does no one have any ideas?

[RupertPupkin]

Originally Posted by dhclose

The whole point of PXE is to be automatic.

Not necessarily. For example, I was forced into a PXE install because the laptop's DVD drive didn't work.

If you want to see what problems occurred during the install -- and obviously there had to have been some or else your machine would be booting properly -- then try an interactive PXE install on the machine to see what happens. Why waste time speculating on what might be the cause of the problem when you can take the guesswork out of it and find out for sure? In the 2 days since your first post you could've done that several times by now.

[dhclose]

Well, I guess I don't know what you mean by an interactive PXE. Do you mean it will work the same as booting from an install DVD?

PXE works from a kickstart file. What do I put in that to make the installation interactive? How can I be sure that the interactive session duplicates the automatic one?

[RupertPupkin]

I used the instructions here to make my main F17 desktop a PXE installation server. It required a bit of configuration but I didn't have to create any kickstart files. Then when I booted my laptop I changed the boot order so that it would try a PXE boot first. After a few seconds the laptop was able to find my PXE installation server and the install started after the necessary files were downloaded over the network from the server. Actually that's where I had my only problem, due to not having enough RAM to load the LiveOS/squashfs.img file from the server. I added more RAM and then it worked.

At the point where the install was ready to start I had the option of running it interactively from the laptop itself or from another machine via VNC. I did it through VNC from my main machine (the server). For you, that is probably where you could start going through the steps indicated in your kickstart file. If it bombs on one of the steps then you'd know what the problem is. If everything works and the machine boots OK, then that would mean that there is likely some syntax error in your kickstart file, or something out of date that no longer works in F17. Either way, you'd have at least some idea of what the problem is. If you posted your kickstart file here maybe someone could spot something wrong with it.

[dhclose]

Originally Posted by RupertPupkin

For you, that is probably where you could start going through the steps indicated in your kickstart file. If it bombs on one of the steps then you'd know what the problem is. If everything works and the machine boots OK, then that would mean that there is likely some syntax error in your kickstart file, or something out of date that no longer works in F17. Either way, you'd have at least some idea of what the problem is. If you posted your kickstart file here maybe someone could spot something wrong with it.

There are basically no steps aside from the disk formatting. It would be very tedious indeed to manually select exactly the same list of packages as specified to kickstart.

The kickstart file is attached. The packages listed are exactly those which are installed on another identical laptop, one installed from the DVD and then customized. I have removed my %post section since the problem occurs even without that present. If I can get the basic installation working, I'm sure I can get %post working.

When running this kickstart, PXE reaches the last package to be installed, then leaves its name on the screen for about 15 minutes during which the disk activity light is nearly constant. Then it tells me that it is installing the bootloader and then reboots. What it is doing during those 15 minutes is a mystery. I've looked around with Ctrl-Alt-F[12345] and don't see any errors. I don't even see any updates during that period. Most of those screens stop updating when package installation starts.

Well, there is one error reported. Some of these laptops have a defective touch screen which continuously generates touch events at random locations, even when no one is nearby. (They are being replaced.) The one I'm using to test has that problem. Before package installation starts, one of the Fx screens reports that the touch event queue has overflowed and that touch events are being discarded. When I get a laptop without this hardware bug, I can prove if that bug is significant. Although I can't be sure, that seems a long shot to me.

[RupertPupkin]

Hmm, maybe your LVM setup didn't get created properly. Did you ever check the output of df to see it it matched what was in your kickstart file?

I notice you have this line in your ks file:

clearpart --linux --drives=sda --initlabel

According to this documentation the --initlabel option only works if the --all option is specified. You do use --all in the clearpart command below that one, so maybe add --initlabel to the second clearpart.

Also, your volgroup line

volgroup vg_laptop --pesize=32768 pv.008002

puts the --pesize option before the name of the partition (pv.008002). The documentation seems to indicate that that order should be reversed:

volgroup vg_laptop pv.008002 --pesize=32768

I don't know if that will really matter but it may be worth trying if nothing else works.

And I noticed that this line

logvol swap --name=lv_swap --vgname=vg_laptop --grow --size=1024 --maxsize=10000

doesn't specify --fstype=swap. Again I don't know if it's really necessary.

[dhclose]

Aside from the machine name, the partitioning section is copied exactly from that in the kickstart file Anaconda generated on the laptop installed from DVD. The filesystem layout looks correct to me. But I'll give these ideas a try.