RHEL 4 Kickstart install not seeing virtual floppy

[Jared Middleton]

Ugh, I'm baffled!

Bare metal server (in another country):
* HP ProLiant DL580 G5
* DVD-ROM, but no Floppy Drive
* Integrated Lights-Out (iLO) 2
* Integrated Remote Console and Virtual Floppy

Scenario:
- Booting DVD "Red Hat Enterprise Linux 4 ES, Update 6" burned from .iso, downloaded from RHN
- A default install via manual interactive GUI appears to work, but I really need to install non-interactive using a Kickstart config file.

Attempted:
1) File ks.cfg on a FAT floppy disk in a standalone USB Floppy Drive attached to the server via USB. Boot RHEL DVD and specify at "boot:" prompt: linux ks=hd:fd0:/ks.cfg (also tried linux ks=floppy)

RESULT: install proceeds as-if "ks=" option ignored (or device not found?)

2) File ks.cfg on a FAT floppy disk in local PC's floppy drive, "mount" iLO Virtual Floppy. Boot RHEL DVD and specify at "boot:" prompt linux ks=hd:fd0:/ks.cfg (also tried: linux ks=floppy)

RESULT: install proceeds as-if "ks=" option ignored (or device not found?)


Other potentially relevant details...

When system is waiting at first RHEL GUI install screen (virtual console 7), I toggle to a new console (CTRL-ALT-F2) to look around...

--- Snippets from /tmp/anaconda.log ---
failed to insert /tmp/floppy.ko
...
getting kickstart file
doing kickstart from floppy
no floppy devices
...
no floppy devices found but we'll try fd0 anyway
--- end of anaconda.log snippets ---

--- Snippets from /tmp/syslog ---
<4> ide-floppy driver 0.99.newide
...
<6> inserting floppy driver for 2.6.9-67.EL
<6> floppy drive(s): fd0 is 1.44M
<4> floppy0: no floppy controllers found
--- end of syslog snippets ---

What do I try next?

Any assistance or pointers would be very much appreciated!

-Jared Middleton

[Ivan Ferreira]

You can place the ks file in the network, use one of the following options:

NFS
===
linux ks=nfs:://

HTTP
====
linux ks=http:///

[Matti_Kurkela]

If you have a USB-connected floppy drive, Linux is more likely to understand it using the "usb-storage" driver. The iLO virtual floppy might work the same way.

After the kernel starts up, Linux "sees through" the BIOS-implemented emulation layer that makes both the USB floppy and the iLO Virtual Floppy seem like a traditional floppy device. That emulation layer can fool the bootloaders and DOS, but not much else.
If Windows shows USB-attached floppy drives as A: or B:, it's probably because Windows notices the size of the device and treats it as a floppy based on that.

Neither is the USB/virtual floppy going to be /dev/fd0 in Linux: it's more likely to get the device name /dev/sda.

MK

[Jared Middleton]

Thanks Ivan. HTTP is exactly what i ended up using last night. I located an intranet server here that I could maintain locally and pointed KS= to it... after booting the Linux DVD in the remote server. I wasn't sure it'd work because I didn't know the DHCP situation. Ultimately, the server will run with a fixed IP, not dynamic. But, for install it worked flawlessly. The server got itself on our WAN, found my kickstart file and proceeded to install all specified packages in like 10 minutes time.

I'll toss some points your way because I didn't get back here to post a summary.

I'd like to keep the thread open a little longer because I had much the same problem a few years back setting up another country's server. I find it hard to believe that it's an HP iLO problem, because the issue would be more well known. It seems to me that Red Hat Linux has a problem supporting Virtual Floppy on a bare metal install. Maybe it all works once linux is installed (with drivers, etc.)?
Anyone know?

[Jared Middleton]

Matti, I agree. Your last sentence is the key. I tried many device name combinations on the server install a couple years ago. My "old" notes implied that a real floppy in an external attached USB Floppy Drive worked with ks=hd:fd0/ks.cfg. But, when i tried that combo now, it didn't work. It's certainly possible that I didn't update my notes with the actual working device name combo? The bummer now is: I'm past the point of being able to test additional scenarios :-(

[Uwe Zessin]

I once saw the following specification:

ks=hd:scd0:/ks.cfg

, but I have not tried it myself.

[Matti_Kurkela]

I've successfully used Uwe's ks= line, with the kickstart configuration file placed on an USB memory stick.

To get more information in your situation, save a copy of /proc/partitions when the virtual floppy is not active. Activate it.
If the device is not auto-detected (check dmesg output), manually load usb-storage and/or any other module that might be helpful.
Then compare the current /proc/partitions to the saved version. After this, you will know how the device is classified in Linux environment.

A plan for the future:
When installing a first unit of a new server model, some experimentation may be required.

1.) Activate the iLO virtual floppy feature, then boot from the installation disc.

2.) Switch to the new console (CTRL-ALT-F2, -F3 or whichever contains a shell) and use it to view /proc/partitions, to identify how the iLO virtual media features work in this particular model

3.) Adapt your Kickstart strategies if necessary; update your notes.

4.) Reboot once more, and install using Kickstart.

MK

[Jared Middleton]

I tried a virtual USB stick too, but that didn't get far. Scene: 4GB thumb drive installed in local PC, activated via iLO, reboot remote server. Result: the server hung early in the POST... never got far enough to enter any device path. From web search, another forum hinted that iLO "virtual" USB might not work with such large sizes (4GB)? But, it was the only one I had, so gave up on that approach.

[Jared Middleton]

While I don't know what would happen on bare metal before Linux install, I just confirmed that the iLO Virtual Floppy is auto-recognized as /dev/sda on a fully running RHEL 4 OS. It is visible in cat /proc/partitions as Matti mentioned.

I can toggle the ILO virtual floppy mount button and the remote Linux console dynamically displays brief info about the newly detected SCSI device sda, and dynamically adds/removes an entry in /etc/fstab. How it comes up with the following strange mount point name I don't know... the fstab entry is:
/dev/sda /media/PKBACK__002 vfat pamconsole,noatime,sync,exec,noauto,managed 0 0

I simply had to mount /media/PKBACK__002 to see the contents of my local Windows floppy (A:)

-Jared

[Matti_Kurkela]

The mount point name you saw is probably based on the volume label of the vfat filesystem. You can set it with the "-n" option of mkdosfs, or using the "mlabel" command on an existing filesystem (you'll need to edit /etc/mtools.conf first, to get a "drive letter" for the /dev/sda device).

It seems likely that the iLO virtual floppy is essentially a "USB storage device" at the bare metal level.

To know for sure, examine /proc/scsi/scsi when the virtual floppy is active. You'll see three lines for each SCSI-like device. If the first line related to the virtual floppy says, for example, "Host: scsi1 [...]", the host adapter number is 1. Then do "ls /proc/scsi/*/1": the directory name (in place of the asterisk) will identify the type of the driver, and thus the controller.

My guess is "usb-storage" or "ub": both are drivers for USB storage devices.

MK