Bug#411166: installation-reports: HP ProLiant BL465c G1
Frans Pop wrote:
On Thursday 22 February 2007 13:46, Sven Coenye wrote:
I tried reinstalling the server, but the installer picks up the
remnants of the partitioning scheme even after writing a blank
partition table to the disk. This leads the installer down a different
path so I can't reproduce this one.
If you delete the LVM headers on the partitions, you should be able to
have a "fresh start". You can even use the installer for that: if you
select manual partitioning there will be an option at the top to
configure LVM and that has options to delete LVs and VGs.
Even with that method, I could not reproduce the problem so the "fresh
start" isn't as fresh as with a new server. Here's a step-by-step
account of everything I selected on a new blade this morning:
1. Boot Debian Etch AMD64 Netinst CD through iLO2
2. Select standard installer with Enter
3. Installation language: english
4. Location: Other/Europe/Belgium
5. Keyboard layout: Belgian
6. Primary network interface: eth0 (Broadcom NetXtreme II BCM5706S)
7. Installer finds the PXE DHCP server and configures but finds no default route (this is as expected.)
7a. Do not continue without a default route
7b. Configure network manually: (IP: 10.2.98.2/16, GW: 10.2.1.250, DNS: 10.2.30.1, hostname: intradb02, domain: henriserruysav.be)
8. Partitioning method: Guided - use entire disk and set up LVM
9. Select disk to partition: cciss/c0d0 (73GB Compaq Smart Array)
10. Partitioning scheme: separate /home, /usr, /var, and /tmp partitions
11. Write changes to disk and configure LVM? Yes
13. Select the partition inside the /home logical volume and delete it.
14. Create a new, smaller partition inside the space. This results in the rest becoming unusable.
15. Realize mistake, delete all other partitions.
16. Still wrong. Delete lvm logical partition.
17. End up with Free space all around, but VG and LV still visible.
Note: At this point I realized that the "Configure the LVM" option is not on the menu (or if it is, its someplace I couldn't see on the iLO remote console). Once at this point, there is no way to rearrange the guided config. This is what is different between an out-of-the-box installation and even the "fresh start" reinstallation of the first server.
18. Create a new primary partition and assign it as LVM volume.
Note: now the "Configure the LVM" option appears at the top of the menu.
19. Select Configure the Logical Volume Manager
20. ... partitioning scheme has to be written to disk: Yes
---> Error informing the kernel about modifications to partition /dev/cciss/c0d0p2 -- Device or resource busy. This means Linux won't know about any changes you made to /dev/cciss/c0d0p2 until you reboot -- so you shouldn't mount it or use it in any way before rebooting. Error!!!
20a. Select default option (Ignore)
21. 1 volume group found. Activate it? Yes. This reenters the partitioner with the erased partions resurrected.
21. Select "Configure the LVM" and erase all logical volumes.
22. Recreate the LVs with the desired sizes.
23. Create partitions inside the LVs
24. Write changes to disk
Note: There's another bug here. I got a warning that I had forgotten to assign a mountpoint to one of the LVs. When one answers "Yes" to "Return to the partitioner?" the installer instead continues with the installation. This one can easily be reproduced each time the LVM partitioner is invoked and exited with an unmounted partition.
25. Select and configure network mirror
26. Select profiles to install
27. Install GRUB on master boot record?: Yes
28. Installation complete, rebooting in new system
29. On reboot, the boot process stops with the "Volume group intradb02" not found message and eventually falls into the repair console.
So this sequence does exactly reproduce the condition of the first installation. I'm hoping this helps. I have a few more untouched blades, but we'll have to move them into production soon.
Reply to: