info
Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Anaconda started via USB for network kickstart, GRUB issues
88 views
Skip to first unread message
Thomas Spear
Nov 1, 2013, 9:37:48 AM11/1/13
to
Hi all,
I'm running RHEL 5.8 and have a semi-uncommon setup for installing RHEL for a corporate environment. I've actually recently inherited the environment from an employee who left a few days after I started, so I'm still figuring things out and I'm hoping I'm just running across a misconfiguration somewhere, but I've reported it on the bugzilla anyways just in case.
Here's the problem. When we install RHEL fresh on machines, GRUB does one of a 3 things depending on certain conditions which I'll describe below. Note: only two of the conditions are problems, the last one I will list is how I would like it to operate regardless of the conditions. Our install is done via kickstart, so the majority of it is automated. The only parts that aren't are the initial IP config in text mode, and the partitioning in GUI mode, which we do the latter via VNC.
So, for all cases below, we go through the entire install.
In the first case, GRUB gets written to the superblock of the USB stick, /dev/sdb (hd1) and grub.conf has "root (hd1,0)". This happens when we leave the USB stick in the machine all the way through the install to the reboot. This is noted by the fact that the syslinux bootloader which was on the USB stick is overwritten with GRUB, and I can boot the OS on the hard drive, as if the USB stick was the drive, resulting in me having to re-copy the USB stick backup I made, to the stick, in order to be able to access the installer again. If I do boot the OS at this point, then remove the USB stick, modify the GRUB device map and grub.conf, and then re-do grub-install, it installs properly to the hard drive and I can boot the machine without the USB stick inserted..
In the second case, GRUB gets written to the superblock of the hard drive like it should be, but the grub.conf gets "root (hd1,0)" anyways. This happens when I remove the USB stick during the file copy phase. Presumably, the grub.conf has already been written at this point, but I can't tell from looking at the kickstart script in our satellite, exactly when it is that grub.conf is being written. In this case, if I re-insert the USB stick, I can boot the installer, and if I remove it, then I have to edit the boot commands before booting in order to successfully boot the OS.
In the final case, GRUB gets written to the superblock of the hard drive like is should be, and grub.conf gets "root (hd0,0)" like it should. This happens when I remove the USB stick before partitioning, but after the GUI (vnc) installer has started. This is how I would like it to be every time.
Now, I could write a post-install script to fix grub.conf, device.map, and re-run grub-install, however that wouldn't prevent the problem of the USB stick's bootloader being overwritten.
How can I make anaconda ignore /dev/sdb entirely for the partitioning and installing of GRUB? If I can't, then is this a bug?
I'm running RHEL 5.8 and have a semi-uncommon setup for installing RHEL for a corporate environment. I've actually recently inherited the environment from an employee who left a few days after I started, so I'm still figuring things out and I'm hoping I'm just running across a misconfiguration somewhere, but I've reported it on the bugzilla anyways just in case.
Here's the problem. When we install RHEL fresh on machines, GRUB does one of a 3 things depending on certain conditions which I'll describe below. Note: only two of the conditions are problems, the last one I will list is how I would like it to operate regardless of the conditions. Our install is done via kickstart, so the majority of it is automated. The only parts that aren't are the initial IP config in text mode, and the partitioning in GUI mode, which we do the latter via VNC.
So, for all cases below, we go through the entire install.
In the first case, GRUB gets written to the superblock of the USB stick, /dev/sdb (hd1) and grub.conf has "root (hd1,0)". This happens when we leave the USB stick in the machine all the way through the install to the reboot. This is noted by the fact that the syslinux bootloader which was on the USB stick is overwritten with GRUB, and I can boot the OS on the hard drive, as if the USB stick was the drive, resulting in me having to re-copy the USB stick backup I made, to the stick, in order to be able to access the installer again. If I do boot the OS at this point, then remove the USB stick, modify the GRUB device map and grub.conf, and then re-do grub-install, it installs properly to the hard drive and I can boot the machine without the USB stick inserted..
In the second case, GRUB gets written to the superblock of the hard drive like it should be, but the grub.conf gets "root (hd1,0)" anyways. This happens when I remove the USB stick during the file copy phase. Presumably, the grub.conf has already been written at this point, but I can't tell from looking at the kickstart script in our satellite, exactly when it is that grub.conf is being written. In this case, if I re-insert the USB stick, I can boot the installer, and if I remove it, then I have to edit the boot commands before booting in order to successfully boot the OS.
In the final case, GRUB gets written to the superblock of the hard drive like is should be, and grub.conf gets "root (hd0,0)" like it should. This happens when I remove the USB stick before partitioning, but after the GUI (vnc) installer has started. This is how I would like it to be every time.
Now, I could write a post-install script to fix grub.conf, device.map, and re-run grub-install, however that wouldn't prevent the problem of the USB stick's bootloader being overwritten.
How can I make anaconda ignore /dev/sdb entirely for the partitioning and installing of GRUB? If I can't, then is this a bug?
0 new messages