After Mike Laverick posted his experience with ESX Server and USB flash disks in another thread, I decided to figure out how to make it work.
Here's a quick rundown on setting up a USB flash disk (ie. pen drive, flash key, etc.) to install ESX Server.
Requirements:
\- You will need at least a 1 GB USB flash disk.
\- You must have
both
the ESX 3.x CD-ROM and ISO image handy.Steps Required:
boot.iso
file in the/images
directory on the ESX 3.x CD-ROM. Copyboot.iso
into a temporary directory on your hard drive.boot.iso
file to your USB flash drive. On my PC, it's theE:
drive.isolinux.bin
andupdatecd.cfg
files from the USB flash disk.isolinux.cfg
file on the USB flash disk tosyslinux.cfg
syslinux.cfg
file and add the keyword usbto the end of every line that begins withappend
. Here's what the file should look like when you're done:syslinux
.zip file into another temporary directory on your hard drive.cd
command to navigate into thewin32
directory. For example:cd C:\temp\syslinux-3.36\win32</p>
9. Now, run the syslinux program to apply the boot loader and boot sector to the USB flash drive:</p>
syslinux -s e:
( On my PC, my USB flash drive is mounted as drive E: )
10. Finally, copy the ESX 3.x ISO image onto the flash drive:
+copy c:\iso\esx-3.0.1-32039.iso e:\ +
11. Confirm that your USB flash drive contains the following files:
boot.cat
boot.msg
initrd.img
snake.msg
splash.lss
vmlinuz
syslinux.cfg
esx-3.0.1-32039.iso
12. You're all set! Unplug the USB flash drive, configure the BIOS on your server to boot from USB accordingly, and boot 'er up!
13. The ESX installer will detect the USB device and whatever SCSI / disk controllers you have. When the installer asks you what the installation source will be, choose
Hard Disk
.Paul
Mike these are the exact same problems I am having. I was using your "ESX01" cfg file you previously posted.
Do you have a working cfg file that I could use that has the %post stuff. I really want to make this work!
Please post the working file or PM me.
Thanks so much!!
So I've taken Pauls work and re-written it here:
http://vmprofessional.com/material/usb_install.html
Perhaps these instructions will help you Mike.
Well, i tried those - and I still have the same problem...
I think it must be the USB drive...
Regards
Mike
Mike,
Have you tried a manual install with the ISO on your USB key? If not, it might help with the debugging. I had no issues using the modified hdstg2.img file supplied in this thread.
Add 'askmethod' to your boot options (and remove KS stuff).
I tried this with a couple of keys, the first (a ByteStor device) gave a "Boot Error" but the second (a Samsung device) works ok.
Thanks
Chris
This has been a great post and I am still trying to tweak the USB boot process. I have taken and also created various KS.cfg files. However, during the install process I get the following error
"tmp/anaconda_build//usr/lib/anaconda/text.py, line 425 in run"
I can boot off the USB drive and do an install, but I would also want to make KS. files loadable from the same USB drive. This is a great idea thanks guys for posting it!!
Guys -
Any idea what I am doing wrong? I followed the instructions. with an older SanDisk 1GB cruzer micro. I am using Hp BL460c blade servers. I am also using iLo for console connection. The system will boot to the usb drive fine. I type in "esx text" or just hit and vmlinuz loads, then initrd.img loads. Then it just stops. No error, no nuttin.
Help :o(
I went around and around and around on this with VMware... Don't ask how I got VMware to look at the problem, but I did. It all boils down to how the usb device is not recognized as a true read/write device and is remounted internally by anaconda, so that the USB boot device is seen as a fake CDROM device, again, anaconda is confused, so it then attempts to mount the USB device now read-only, as read-writeable, and thus partitions.py freaks out. Thinking I could side step the issue, I got the USB device to be recognized as a Hard-Disk... well no, due to the device enumeration of type, I hit a break wall again. Dang anaconda continues to see the USB device it booted from as a VALID disk it SHOULD mount as read-write.
This is not technically a bug in partitions.py, but something that partitions.py coding is just not designed to do because it only supports 3 types, CDROM, HD (SCSI or otherwise), and Floppy device types. To correct the issue per VMware, anaconda would have to be significantly changed in several places, starting with the actual code that does the initial device discovery at the hardware analysis level which core to ASM of anaconda, and teaching anaconda what it means to be a 21st centry 2007 tool. This, per VMware, is just not worth the effort, or at least that was the comment at the time.
I'm not even getting to anaconda. This is before the drivers even load. It just sticks at "Loading initrd.img ........."
I went around and around and around on this with
VMware... Don't ask how I got VMware to look at the
problem, but I did. It all boils down to how the usb
device is not recognized as a true read/write device
and is remounted internally by anaconda, so that the
USB boot device is seen as a fake CDROM device,
again, anaconda is confused, so it then attempts to
mount the USB device now read-only, as
read-writeable, and thus partitions.py freaks out.
Thinking I could side step the issue, I got the USB
device to be recognized as a Hard-Disk... well no,
due to the device enumeration of type, I hit a break
wall again. Dang anaconda continues to see the USB
device it booted from as a VALID disk it SHOULD
mount as read-write.
This is not technically a bug in partitions.py, but
something that partitions.py coding is just not
designed to do because it only supports 3 types,
CDROM, HD (SCSI or otherwise), and Floppy device
types. To correct the issue per VMware, anaconda
would have to be significantly changed in several
places, starting with the actual code that does the
initial device discovery at the hardware analysis
level which core to ASM of anaconda, and teaching
anaconda what it means to be a 21st centry 2007 tool.
This, per VMware, is just not worth the effort, or
at least that was the comment at the time.
I think this post is very meaningful/useful... I decide a couple of months ago to give on this esx-on-a-stick approach. I just never got it working, and became more interested in a over-the-network PXE approach which became the UDA...
Regards
Mike
Hi Paul, I'm wondering if you have a patched version of the of the hdstg2.img file for 3.5
Thanks
Mike
Thanx to Mike & Paul (et al.) for all their hard work in posting their info., Just thought I'd share my onw expereinces whilst I'm at a dead end
I'm using a OEM Blade with a P35 (ICH9R) chipset & 2 500GB SATA . It 's the only box avaibale in the firm that is spare & I can utilise for prep on my upcoming VCP Course, and the snads of time are running out quickly!!
- CD-ROM Installation hits a bring wall of 'Device not Fóund '
- I've been meticulous in the USB Postings, along with things such as recreating the ESX ISO with mkisofs via RHEL & not any Windows ISO Product (which leads to the same error above) & arrive at the stage where, whilst the SATA Disk & their partitons are seen, the installation complains of
The ESX Tree does not match your media
It seems that recerating the ISO with the patched hdstg2.img may have created that..I hope, Less I'll order on my own account a SCSI Controller & Card
BTW I've tried ALT-F@ at anaconda ( to get to the shell & investigate if their any problems with how the USB Partitons are recognised )which seems to have been disabled ?
Realised eslewhere that the hdstg2.img version does match the initrd.img
Hence joined the queue with Mike (see previous post)
~ |
Proceeded further with a unmodified ESX 3.5 ISO on the USB, now awaiting a HCL-compatible NIC
Using Pauls modified .img file is incompatiable with 3.5. I've been trying to insert Pauls modified partitions.pyc file into the hgstg2.img from 3.5. No luck. Paul, if you get a chance could you explain how you modified the .img file.
Regards
Mike F.
Mike
See my previous post, I discarded esx-custom.iso & replaced with the original ESX 3.5 ISO & have proceeded further
Stevie (ms.)
I tried that but the partitions.pyc file is still complaining.
Tried the 64607 Build of ESX (28th February Release) ISO on the USB, failed with the familiar error of ' The Tree not matching the media Type'
Sucess: Managed to install the 64607 build over SCSI (70GB U160 + Adaptec29320-R via auction) via HTTP...no other method including CD-ROM would worked
Almost Sucess: Installed over a existing Fedora 7 SATA Disk, but would not GRUB. Manually updated GRUB after mounting ESX's /boot, but Fedora's GRUB, didn't recognise VMFS , so reverting back to SCSI - ...
Good Luck All
Tried the VI 3.5 (64607 Build of ESX, 28th February Release) ISO on the USB, failed with the familiar error of ' The Tree not matching the media Type.
Used Paul's hdstg2.img file.
Have anybody managed to get a USB install of VI 3.5 working ?
Jan Erik Holo
I haven't gotten it to work. Paul's hdstg2.img file is for 3.01. if you loop mount the img you will see a lot of python files in the anaconda directory. They are not compatible with 3.5