anaconda fails to mount a newly-created root partition, and so attempts to install the OS onto the RAM disk, causing an error about there not being enough disk space to proceed with the install. How Reproducible: With a ks.cfg containing: part / --fstype ext3 --size 9000 --grow anaconda successfully creates the new / partition, but then complains it isn't big enough. This is probably because the ext3.o kernel module is missing from the relevant boot image (in this case, redhat/linux/8.0/en/os/i386/images/pxeboot/initrd.img) Please note that the correct fix for this is not just to add ext3.o to the image, but also to add a check for mount failure to anaconda, so that if the mount of the new / partition fails, the error message says this rather than the misleading report that the partition was too small.
The ext3 module doesn't have to be on the first stage; it is in the second stage moduleball. Could you check if the ext3 module gets loaded then and if not, what errors are given on tty3. Also, could you attach your complete ks.cfg?
What do you mean by the first stage and second stage module ball exactly? I was under the impression that the first kernel modules the install sees are in pxeboot/initrd.img, and no more are encountered before the partitioning happens, hence the need for ext3.o to be in this initrd.img. I don't remember seeing any attempts to insmod ext3.o on tty3 at the time, although it's possible that the attempt could have just scrolled past, since the kernel loses all scrollback every time you switch ttys. I should be able to verify this tomorrow. If I added ext3.o to modules.cgz within pxeboot/initrd.img, wouldn't this fix the problem? [ks.cfg to be attached shortly]
Created attachment 80524 [details] kickstart config
Bit puzzled why this has been changed to NEEDINFO. Do you need any more info from me?
It appears that you were right, and this problem is more complex than I first thought. The same problem has just cropped up with another ks.cfg. Using the root shell on tty2 reveals that it *has* successfully mounted the new ext3 root partition on /mnt/sysimage, and yet it still claims that there is not enough space! The exact error message is: You don't appear to have enough disk space to install the packages you've selected. You need more space on the following file systems: Mount Point Space Needed / 1683 M OK However, as you will see from the attached outputs of the mount and df commands run on tty2, this is most definitely untrue!
Created attachment 82404 [details] Output of df run on tty2 after bogus anaconda error occurred
Created attachment 82405 [details] Output of mount run on tty2 after bogus anaconda error occurred
Ahhhhhhhhh ... ! When it said "space needed: 1683MB", it meant 1683 *more* MB, not 1683MB in total. And full install of RedHat 8.0 is close on 5GB! (When did it get so huge?!) So the original problem stands - there *is* something funny with ext3, but you can ignore today's comments and attachments. Could it be something to do with the fact that I was using RedHat 8.0's anaconda to perform an install of Advanced Server 2.1?
There is NO way that will work. The modules on the second stage (including the ext3 module) won't match the version of your running kernel and so can't be loaded if you do this.