Created attachment 929322 [details] /tmp and /var/log from unsuccessful installation Description of problem: F20 x86_64 installation with kickstart without describing disk partitioning (wanting to manually assign pre-split disks and RAID arrays created). It was to be small NAS composed of five 2TB drives in two RAID5 MD arrays and sixth solo 80 GB disc on which the system should be. This 80 GB disk had pre-prepaged these GPT partitions: - 1MB sda1 type ef02 (should be BIOS boot) - 400 MB sda2 type 8300 (should be ext4 /boot) - 4 GB sda3 type 8200 (should be swap) - 32 GB sda4 type 8300 (should be ext4 root /) - rest (?35 GB?) sda5 type 8300 (should be ext4 /home) After anaconda start to initial graphic window, I chose disc partitioning, there was all these 6 disks selected, in next window I selected standard partitions and manual partitioning. In manual partitioning window under "Unknown" tab I saw my sda1-sda5 partitions + two RAID5 partitions, and could reformat them as I wanted. After updating all partitions, these ALL moved from "Unknown" tab to "New Fedora 20 Installation" tab. Then I clicked "Done" and came back to 'INSTALLATION SUMMARY' screen. But "INSTALLATION DESTINATION" item had warning triangle with a message 'Error checking storage configuration' and after clicking on it and following '...Click for details', I saw window with: "Your BIOS-based system needs a special partition to boot from GPT disk label. To continue, please create a !MB 'biosboot' type partition." And when continuing clicking on "Modify Storage Layout", on 'MANUAL PARTITIONING' window I saw my sda1 as 'BIOS Boot' _UNDER_ 'Unknown' tab. And here I ended, I could not in any way to get this sda1 up under the 'New Fedora 20 Installation' tab, I must quit installation. And now strange thing: when I disconnected five 2 TB drives with pre-prepared RAIDs and leave only this 80GB sda, then I was able complete installation successfuly, without problems, and 'biosboot' partition now was accepted by anaconda. Version-Release number of selected component (if applicable): 20.25.15-1 How reproducible: This was on AMD 2-core system with 4 GB DDR3 RAM, maybe HW does not matter here, and it appears on any other systems. Additional info: attached is contens of /tmp/ and /var/log/ directories from unsuccessful installation. There is only removed /tmp/yum.cache/*, because it perhaps isn't important and it is 21+ MB big.
Created attachment 929324 [details] /tmp and /var/log from successful installation (sda only) content of /tmp/ and /var/log/ (again without /tmp/yum.cache/*) from successful installation, where in F20 installation time was connected only sda 80GB disk. Content is from the moment before clicking the 'Begin installation' button.
This problem seems be not related to MD RAIDs, but to existence two (+?) physical disks. Now I tried another F20 x86_64 installation from kickstart, partitions on two disks, /dev/sda (biosboot, /boot, swap, / with rest of the system) and sdb (/home and /mnt/data), were prepared before with gdisk. Installation began without problems, but then crashes in "Starting package installation process" phase due to missing one dependent package in my repo. After zeroing beginnings of partitions (see my commant to bug 1126064 https://bugzilla.redhat.com/show_bug.cgi?id=1126064#c2 ), I was able to manually arrange partitions (determine their types, labels, formats, mount points). But after clicking to "Done", in 'INSTALLATION SUMMARY' screen I see message 'Error checking storage configuration' and as reason "Your BIOS-based system needs a special partition to boot from GPT disk label. To continue, please create a 1MB 'biosboot' type partition." - although this partition was there. I find that I do not have to disconnect sdb physically, but just uncheck it on window where are physical disks displayed is sufficient. And the I was able successfully complete the installation.
You need to reformat the biosboot partition.