Documentation Installer Beat
Anaconda is the name of the Fedora installer. This section outlines issues related to Anaconda and installing Fedora 10.
Anaconda asks if it should verify the installation medium when Install or upgrade an existing system is selected during boot from an installation-only media.
For Fedora Live media, press any key during the initial boot countdown to display a boot option menu. Select Verify and boot to perform the media test. Installation media can be used to verify Fedora Live media. Anaconda asks during the mediacheck if you want to check any other disc than the one Anaconda is running from. To test additional media, select
eject to eject the inserted medium, then replace it with the medium you want to test.
Perform this test for any new installation or Live medium.
The Fedora Project strongly recommends that you perform this test before reporting any installation-related bugs. Many of the bugs reported are actually due to improperly-burned CD or DVDs.
In rare cases, the testing procedure may report some usable discs as faulty. This result is often caused by disc writing software that does not include padding when creating discs from ISO files.
Another reason for a failure during installation is faulty memory. To perform memory testing before you install Fedora, press any key to enter the boot menu, then select Memory Test. This option runs the Memtest86 standalone memory testing software in place of Anaconda. Memtest86 memory testing continues until you press the [Esc] key.
Fedora 10 supports graphical FTP and HTTP installations. However, the installer image must either fit in RAM or appear on local storage, such as the installation DVD or Live Media. Therefore, only systems with more than 192MiB of RAM or that boot from the installation DVD or Live Media can use the graphical installer. Systems with 192MiB RAM or less fall back to using the text-based installer automatically. If you prefer to use the text-based installer, type
linux text at the
Changes in Anaconda
- NetConfig in NM -- Anaconda is now using NetworkManager for configuring network interfaces during installation. The previous backend tool was
libdhcp, which was a replacement for
libpump. Anaconda uses NetworkManager by communicating with it via D-Bus during installation. The move to NetworkManager in Anaconda is still ongoing, but the bulk of existing functionality has been retained. NetworkManager is enabled by default on newly installed systems, so moving to NetworkManager in Anaconda allows the installer to use the same network management tool that the final system uses. The move to NetworkManager brings some changes, most notably the removal of the network interface configuration screen in Anaconda. You are no longer asked to verify the network settings during installation. The screen now simply prompts for the hostname. The settings used during installation are written to the system.
- When using the
netinst.isoto boot the installer, Anaconda defaults to using the Fedora mirrorlist URL as the installation source. The method selection screen no longer appears by default. If you do not wish to use the mirrorlist URL, either add
repo=<your installation source>or add
askmethodto the installer boot parameters. The
askmethodoption causes the selection screen to appear as it did in previous releases. To add boot parameters, press the [Tab] key in the initial boot screen and append any new parameters to the existing list. For more information, see the
stage2=descriptions at http://fedoraproject.org/wiki/Anaconda/Options
Installation Related Issues
- When PXE booting and using a .iso file for the installation media via NFS you are now required to add
method=nfsiso:server:/pathto the command line.
IDE device names
/dev/hdX on i386 and x86_64 for IDE drives changed to
/dev/sdX in Fedora 7. If you are upgrading from an earlier version than Fedora 7, you need to research about the importance of labeling devices for upgrades and any partition limitations.
Not all IDE RAID controllers are supported. If your RAID controller is not yet supported by
dmraid, you may combine drives into RAID arrays by configuring Linux software RAID. For supported controllers, configure the RAID functions in the computer BIOS.
Multiple NICs and PXE installation
Some servers with multiple network interfaces may not assign eth0 to the first network interface as BIOS knows it, which can cause the installer to try using a different network interface than was used by PXE. To change this behavior, use the following in
pxelinux.cfg/* config files:
IPAPPEND 2 APPEND ksdevice=bootif
The configuration options above causes the installer to use the same network interface as BIOS and PXE use. You can also use the following option:
This option causes the installer to use the first network device it finds that is linked to a network switch.
Refer to http://fedoraproject.org/wiki/DistributionUpgrades for detailed recommended procedures for upgrading Fedora.
SCSI driver partition limits
Whereas older IDE drivers supported up to 63 partitions per device, SCSI devices are limited to 15 partitions per device. Anaconda uses the
libata driver in the same fashion as the rest of Fedora, so it is unable to detect more than 15 partitions on an IDE disk during the installation or upgrade process.
If you are upgrading a system with more than 15 partitions, you may need to migrate the disk to Logical Volume Management (LVM). This restriction may cause conflicts with other installed systems if they do not support LVM. Most modern Linux distributions support LVM and drivers are available for other operating systems as well.
Disk partitions must be labeled
A change in the way that the linux kernel handles storage devices means that device names like
/dev/sdX may differ from the values used in earlier releases. Anaconda solves this problem by relying on partition labels or UUIDs for finding devices. If these are not present, then Anaconda presents a warning indicating that partitions need to be labelled and that the upgrade can not proceed. Systems that use Logical Volume Management (LVM) and the device mapper usually do not require relabeling.
To check disk partition labels
To view partition labels, boot the existing Fedora installation, and enter the following at a terminal prompt:
Confirm that each volume line in the list has a
LABEL= value, as shown below:
/dev/hdd1: LABEL="/boot" UUID="ec6a9d6c-6f05-487e-a8bd-a2594b854406" SEC_TYPE="ext2" TYPE="ext3"
To set disk partition labels
For ext2 and ext3 partitions without a label, use the following command:
su -c 'e2label /dev/example f7-slash'
For a VFAT file system use
dosfslabel from the
dosfstools package, and for NTFS file system use
ntfslabel from the
ntfsprogs package. Before rebooting the machine, also update the file system mount entries, and the GRUB kernel root entry.
Update the file system mount entries
If any file system labels were added or modified, then the device entries in
/etc/fstab must be adjusted to match:
su -c 'cp /etc/fstab /etc/fstab.orig' su -c 'gedit /etc/fstab'
An example of a mount by label entry is:
LABEL=f7-slash / ext3 defaults 1 1
Update the grub.conf kernel root entry
If the label for the / (root) file system was modified, the kernel boot parameter in the grub configuration file must also be modified:
su -c 'gedit /boot/grub/grub.conf'
A matching example kernel grub line is:
kernel /vmlinuz-2.6.20-1.2948.fc6 ro root=LABEL=f7-slash rhgb quiet
Test changes made to labels
If partition labels were adjusted, or the
/etc/fstab file modified, then boot the existing Fedora installation to confirm that all partitions still mount normally and login is successful. When complete, reboot with the installation media to start the installer and begin the upgrade.
Upgrades versus fresh installations
In general, fresh installations are recommended over upgrades. This is particularly true for systems that include software from third-party repositories. Third-party packages remaining from a previous installation may not work as expected on an upgraded Fedora system. If you decide to perform an upgrade anyway, the following information may be helpful:
Before you upgrade, back up the system completely. In particular, preserve
/home, and possibly
/usr/local if customized packages are installed there. You may want to use a multi-boot approach with a "clone" of the old installation on alternate partition(s) as a fallback. In that case, create alternate boot media, such as a GRUB boot floppy.
After you complete the upgrade, run the following command:
rpm -qa --last > RPMS_by_Install_Time.txt
Inspect the end of the output for packages that pre-date the upgrade. Remove or upgrade those packages from third-party repositories, or otherwise deal with them as necessary. Some previously installed packages may no longer be available in any configured repository. To list all these packages, use the following command:
su -c 'yum list extras'
Kickstart HTTP Issue
When using a Kickstart configuration file via HTTP, kickstart file retrieval may fail with an error that indicates the file could not be retrieved. Click the OK button several times without making modifications to override this error successfully. As a workaround, use one of the other supported methods to retrieve Kickstart configurations.