Upgrading from EOL Fedora using package manager

Click here to set up this language box

Read Upgrading Fedora using package manager as well
This page contains notes on issues you might encounter running yum update between specific EOL Fedora releases, and variant general instructions for upgrading from very old Fedora releases (Fedora 11 and earlier). Upgrading Fedora using package manager contains general notes and instructions on performing upgrades using the bare package manager. Please be sure to read Upgrading Fedora using package manager for an overview of the entire process as well as the release-specific notes for upgrades between the oldest supported release and the current release.

All upgrades from Fedora < 17: /usr move

First install the new Fedora 17 gpg key

rpm --import https://fedoraproject.org/static/1ACA3465.txt

There was a major change in Fedora 17 called the '/usr move'. Fedora 17 located the entire base operating system in /usr. From Fedora 17 onwards, the directories /bin, /sbin, /lib, /lib64 are only symlinks:

/bin → /usr/bin
/sbin → /usr/sbin
/lib → /usr/lib
/lib64 → /usr/lib64

Some reasoning behind this change is outlined here.

In order to upgrade to Fedora 17 or later, you must first upgrade to Fedora 16. Then, you must upgrade to Fedora 17 after performing some manual steps to convert the current system to match the layout of Fedora 17 and later. These steps are documented in the Fedora 17 instructions. After that, you can continue upgrading to a later release as usual.

Some packages in Fedora 17 are carrying an RPM dependency guard, which will make sure they can only be installed when /bin, /sbin, /lib, /lib64 are symlinks and not directories like in Fedora 16 and older.

The installed system’s base filesystem layout can not be safely altered, while the system itself is running on top of it. Dracut, the initramfs used to find and mount the root filesystem, can be instructed to convert the filesystem to match Fedora 17's expectations.

If your system has a split-off /usr, a separate mount point, the dracut /usr mount conversion logic might not work. If /usr resides on the net, then you should add "rd.neednet=1" and the network settings like "ip=dhcp" on the kernel command line.

/usr on iSCSI, FCoE, NBD also is supported, as long as “netroot=...” is specified on the kernel command line for these disks (see man dracut.kernel(7)). If you have /usr on LVM, MD raid or DM raid, make sure the kernel command line has either all settings like {{{1}}} to ensure the /usr device is accessible in dracut or just remove all restrictions like rd.lvm..., rd.md..., rd.dm.... Either way, you should probably use anaconda to upgrade, if you are experiencing problems with a separate /usr.

You will have to manually convert /var/run and /var/lock to a symbolic link.

# mv -f /var/run /var/run.runmove~
# ln -sfn ../run /var/run
# mv -f /var/lock /var/lock.lockmove~
# ln -sfn ../run/lock /var/lock

Here are the steps to prepare your system, to convert it, and to be able to continue updating your installed system with yum:

Download and install the most recent dracut packages:

# yum update dracut

You should at least have dracut-013-22.fc16.

Turn off any "hostonly" settings in /etc/dracut.conf*, if you turned on "hostonly".

Update the installed initramfs image for your current kernel, and instruct dracut to include the dracut module to convert your current filesystem:

# dracut --force --add convertfs

If your system has a split-off /usr, a separate mount point, and you don't know the kernel command line parameter to add, you can also try (dracut tries to generate them internally), but you have to install at least dracut-018.

# dracut -H --force --add convertfs

If dracut detects ‘rd.convertfs’ on the kernel command line at bootup, it starts the filesystem conversion of the root filesystem. If it is already converted, it will just do nothing.

dracut will by default update the initramfs for the currently running kernel. Make sure that no new kernel has been installed since last boot and that you really will boot into the updated initramfs.

Change the following kernel commandline parameter directly in the bootloader menu, which is shown during bootup, or edit the line in /etc/grub*.cfg to remove ro and rhgb and append rw rd.info rd.convertfs enforcing=0

Explanation of the options:

- remove “ro” (read only)
- append “rw” (read write) to let dracut mount your root filesystem writeable
- remove “rhgb” (Red Hat graphical boot) to disable the graphical bootsplash
- append “rd.info” to get a more verbose output from dracut
- append “rd.convertfs” to enable the /usr-move conversion script in dracut
- append “enforcing=0” to disable SELinux enforcement

During bootup, dracut will now convert your filesystem, and /lib, /lib64, /bin and /sbin should then all be symbolic links to the corresponding directories in /usr.

After the conversion, the system needs to be immediately updated to Fedora 17. No packages from Fedora 16 or Fedora 15, or older rawhide packages must be installed any more. Make sure to disable any Fedora 15 and Fedora 16 repositories in yum!

Any files with conflicting names, which the conversion could not resolve, will be backed up to files named *.usrmove~ residing in /usr/lib, /usr/lib64, /usr/bin and /usr/sbin.

Verify that dracut really completed the conversion. The log messages, which dracut has generated during bootup, can be retrieved with:

# dmesg | grep dracut

After a successful conversion, revert the changes made to the kernel command line in the bootloader config file /etc/grub*.cfg.

Then run,

# rm -f /var/lib/rpm/__*
# rpm --rebuilddb
# yum --releasever=17 update rpm
# rm -f /var/lib/rpm/__*
# rpm --rebuilddb
# yum --releasever=17 --disableplugin=presto distro-sync
# fixfiles onboot  

After upgrading, all should be set and done. You can then proceed with upgrading to a later release.

All upgrades from Fedora < 21: Fedora 'flavors'

Aside from all other information on this page, any time you upgrade from Fedora 20 or earlier to a currently-supported Fedora release, you may install a Fedora 'flavor' package (and package group). With the Fedora 21 release, the concept of 'flavors' was introduced, with the Server, Workstation, and Cloud flavors. See Fedora.next for more details. To have your system considered a Fedora Server, Fedora Workstation or Fedora Cloud system, you can run:

dnf install fedora-release-(flavor)
dnf groupinstall (flavor)-product-environment

If you do not do this, your system will be considered a non-flavor install, much like an install from e.g. the KDE or Xfce live images would be.

Upgrading from Fedora 12 and later

The general procedure outlined on the Upgrading Fedora using package manager page should apply to upgrades from Fedora 12 and later, with the caveat that up to Fedora 21, in all commands, you should replace dnf with yum.

For special instructions for upgrading from Fedora 11 and earlier, and the release-specific notes for those releases, see the Upgrading from Fedora 11 and earlier section later on this page.

Release specific notes for EOL releases Fedora 12 and later

Fedora 21 -> Fedora 22

No special instructions. Follow the above instructions.

Fedora 20 -> Fedora 21

No special instructions. Follow the above instructions. Also see the 'flavors' section.

Many people report (e.g. bugs 1035326, 1044184, 1002331) that after upgrade some files have incorrect SELinux context. It is recommended that you run:

 # restorecon -R /

Fedora 19 -> Fedora 20

No special instructions. Follow the above instructions.

Fedora 18 -> Fedora 19

Incompatible systemd cgroups hierarchy layout
systemd in F19 lays out its cgroups hierarchy differently than in previous Fedora releases.

The new hierarchy has some advantages, but it is incompatible with the old one. No live conversion of the hierarchy is performed during the package upgrade, so the upgraded systemd will not understand the previous state of the system correctly. This is known to affect the tracking of user sessions by systemd-logind (bug #962983).

Expect breakage in active user sessions and make sure to reboot soon after performing the upgrade. If your screen locks during the upgrade, you may not be able to log back in due to the cgroup changes. Running the upgrade in a screen(1) session has been reported as a workaround (re-attach from a vt).

Install the new Fedora 19 gpg key and upgrade:

# rpm --import https://fedoraproject.org/static/FB4B18E6.txt
# yum update yum
# yum clean all
# yum --releasever=19 --disableplugin=presto distro-sync
# 

If you use Fedora ARM, run these commands instead:

# rpm --import https://fedoraproject.org/static/BA094068.txt
# yum update yum
# yum clean all
# yum --releasever=19 distro-sync

Fedora 17 -> Fedora 18

Recommended Upgrade Method (Fedora 17 and newer)
Upgrading with FedUp is the recommended way to upgrade your system!Please check How Can I Upgrade My System with FedUp?

Note: One user has reported problems trying to upgrade an Intel Mac UEFI installation using this method, including the manual bootloader migration. Read more

1. Install the new Fedora 18 gpg key: su -c 'rpm --import https://fedoraproject.org/static/DE7F38BD.txt'
2. If you are using SELinux in Enforcing mode, make sure your selinux-policy is up to date: su -c 'yum update selinux-policy'
3. Upgrade all packages:
   1. su -c 'yum update yum'
   2. su -c 'yum clean all'
   3. su -c 'yum --releasever=18 --disableplugin=presto distro-sync' It is recommended that updates be enabled for this step, for best results.
4. Rebuild rpm database: su -c 'rpm --rebuilddb', or rpm -qa will not work due to an upgrade of rpm

If you used an outdated selinux-policy package in Enforcing mode during the upgrade, you may notice errors in the yum output when several packages attempt to create users and/or groups, and after the upgrade you may have problems related to these packages, including issues with logging in via GDM (you may just see a spinning cursor) and/or performing actions requiring administrative privileges. This is a result of bug #844167. If you have this problem, you should re-install the affected packages with su -c 'yum reinstall (packagenames)', and then reboot. Affected packages may include libvirt-daemon and polkit: su -c 'yum reinstall libvirt-daemon polkit'

Due to Features/DisplayManagerRework, the upgrade may leave you without a display manager enabled. To solve this problem, use su -c 'systemctl enable yourdm.service', replacing yourdm with the display manager you intend to use, e.g. gdm or kdm.

Fedora 17

First install the Fedora 17 package signing key:

rpm --import https://fedoraproject.org/static/1ACA3465.txt

There was a major change in Fedora 17 called the '/usr move'. Fedora 17 located the entire base operating system in /usr. From Fedora 17 onwards, the directories /bin, /sbin, /lib, /lib64 are only symlinks:

/bin → /usr/bin
/sbin → /usr/sbin
/lib → /usr/lib
/lib64 → /usr/lib64

Some reasoning behind this change is outlined here.

In order to upgrade to Fedora 17, you must first upgrade to Fedora 16.

Some packages in Fedora 17 are carrying an RPM dependency guard, which will make sure they can only be installed when /bin, /sbin, /lib, /lib64 are symlinks and not directories like in Fedora 16 and older.

The installed system’s base filesystem layout cannot be safely altered while the system itself is running on top of it. Dracut, the initramfs used to find and mount the root filesystem, can be instructed to convert the filesystem to match Fedora 17's expectations.

If your system has a split-off /usr, a separate mount point, the dracut /usr mount conversion logic might not work. If /usr resides on the net, then you should add "rd.neednet=1" and the network settings like "ip=dhcp" on the kernel command line.

/usr on iSCSI, FCoE, NBD also is supported, as long as “netroot=...” is specified on the kernel command line for these disks (see man dracut.kernel(7)). If you have /usr on LVM, MD raid or DM raid, make sure the kernel command line has either all settings like {{{1}}} to ensure the /usr device is accessible in dracut or just remove all restrictions like rd.lvm..., rd.md..., rd.dm.... Either way, you should probably use anaconda to upgrade, if you are experiencing problems with a separate /usr.

You will have to manually convert /var/run and /var/lock to a symbolic link.

# mv -f /var/run /var/run.runmove~
# ln -sfn ../run /var/run
# mv -f /var/lock /var/lock.lockmove~
# ln -sfn ../run/lock /var/lock

Here are the steps to prepare your system, to convert it, and to be able to continue updating your installed system with yum:

Download and install the most recent dracut packages:

# yum update dracut

You should at least have dracut-013-22.fc16.

Turn off any "hostonly" settings in /etc/dracut.conf*, if you turned on "hostonly".

Update the installed initramfs image for your current kernel, and instruct dracut to include the dracut module to convert your current filesystem:

# dracut --force --add convertfs

If your system has a split-off /usr, a separate mount point, and you don't know the kernel command line parameter to add, you can also try (dracut tries to generate them internally), but you have to install at least dracut-018.

# dracut -H --force --add convertfs

If dracut detects ‘rd.convertfs’ on the kernel command line at bootup, it starts the filesystem conversion of the root filesystem. If it is already converted, it will just do nothing.

dracut will by default update the initramfs for the currently running kernel. Make sure that no new kernel has been installed since last boot and that you really will boot into the updated initramfs.

Change the following kernel commandline parameter directly in the bootloader menu, which is shown during bootup, or edit the line in /etc/grub*.cfg to remove ro and rhgb and append rw rd.info rd.convertfs enforcing=0

Explanation of the options:

- remove “ro” (read only)
- append “rw” (read write) to let dracut mount your root filesystem writeable
- remove “rhgb” (Red Hat graphical boot) to disable the graphical bootsplash
- append “rd.info” to get a more verbose output from dracut
- append “rd.convertfs” to enable the /usr-move conversion script in dracut
- append “enforcing=0” to disable SELinux enforcement

During bootup, dracut will now convert your filesystem, and /lib, /lib64, /bin and /sbin should then all be symbolic links to the corresponding directories in /usr.

After the conversion, the system needs to be immediately updated to Fedora 17. No packages from Fedora 16 or Fedora 15, or older rawhide packages must be installed any more. Make sure to disable any Fedora 15 and Fedora 16 repositories in yum!

Any files with conflicting names, which the conversion could not resolve, will be backed up to files named *.usrmove~ residing in /usr/lib, /usr/lib64, /usr/bin and /usr/sbin.

Verify that dracut really completed the conversion. The log messages, which dracut has generated during bootup, can be retrieved with:

# dmesg | grep dracut

After a successful conversion, revert the changes made to the kernel command line in the bootloader config file /etc/grub*.cfg.

Then run,

# rm -f /var/lib/rpm/__*
# rpm --rebuilddb
# yum --releasever=17 update rpm
# rm -f /var/lib/rpm/__*
# rpm --rebuilddb
# yum --releasever=17 --disableplugin=presto distro-sync
# fixfiles onboot  

Fedora 16

First install the Fedora 16 package signing key. You may wish to verify this package against https://fedoraproject.org/keys and the fedora ssl certificate.

rpm --import https://fedoraproject.org/static/A82BA4B7.txt

Next run chkconfig --list and note the enabled services; you will need to re-enable these with systemctl enable xxxxx.service after you reboot, as the sysvinit settings aren't propagated into systemd. See release notes for more details.

Upgrade all packages with

yum update yum
yum clean all
yum --releasever=16 --disableplugin=presto distro-sync

Bootloader change
After the upgrade, you will have the grub2 and grub-efi packages installed and the grub package removed. However, grub will still be installed to the MBR and booting should still work. You can follow the instructions below to switch to grub2 if you choose. If installed, the firstaidkit-plugin-grub package may prevent the upgrade; do simply remove that package if that is the case.

If your system uses a BIOS, or you installed Fedora via BIOS emulation mode on an EFI system (not native EFI mode), you can switch to Fedora 16's supported grub2 bootloader with the following instructions. If your system was installed by native EFI boot, do not switch to grub2, as its EFI support is still unreliable. Fedora 16's supported bootloader for native EFI installations is still grub-legacy, so you should simply continue to use the system without making any special changes to the bootloader configuration.

To switch to grub2, run the command su -c '/sbin/grub2-mkconfig -o /boot/grub2/grub.cfg', then proceed as described above with reinstalling the bootloader, but call grub2-install /dev/XXX instead of grub-install /dev/XXX.

Known you upgrade -specific issues (for common problems, see references above):

• Bug 743022 - F15->F16 yum update fails with IMSM (BIOS) raid

Fedora 15

First install the new fedora 15 gpg key. You may wish to verify this package against https://fedoraproject.org/keys and the fedora ssl certificate.

rpm --import https://fedoraproject.org/static/069C8460.txt

Upgrade all packages with

yum update yum
yum clean all
yum --releasever=15 --disableplugin=presto distro-sync

• Do not run this from within an X terminal. Testing shows that X might hang while updating bitmap font packages.
• There exist .drpms, but they don't match, due to a format change, so better disable the presto plugin by adding the option "--disableplugin=presto" (without quotes) when running yum.
• The F15 screen client is not capable of attaching to F14 screen sessions. Thus, if you want to run the upgrade under screen, you should either upgrade screen itself in a separate operation or make a separate copy of screen to use throughout the process.
• mysql 5.5.20 shipped with F15 uses InnoDB as default storage engine. After upgrading, mysqld could refuse to start-up with error Unknown/unsupported storage engine: InnoDB if argument skip-innodb is given on command line or configuration file /etc/my.cnf. Workarounds are removing the line (InnoDB will be run as default engine), or adding default-storage-engine command options specifying some other storage engine.

VirtualBox guest upgrades

The steps above work perfectly for upgrading a Fedora 14 guest to Fedora 15, but you also need to remove the Guest Additions. If you forget, F14 -> F15 upgrades will seem to fail after the first reboot. If that happens, log in to the console with CTRL+ALT+F2 and reinstall the guest additions manually:

mount /dev/cdrom /media
# if /dev/cdrom does not exist, try:
# mount /dev/sr0 /media
/bin/sh /media/VBoxLinuxAdditions.run
reboot

Fedora 14

First install the new fedora 14 gpg key. You may wish to verify this package against https://fedoraproject.org/keys and the fedora ssl certificate.

rpm --import https://fedoraproject.org/static/97A1071F.txt

Upgrade all packages with

yum update yum
yum clean all
yum --releasever=14 distro-sync

• If using VirtualBox from the Oracle repository, you must remove the VirtualBox-3.1 package before upgrading. After the upgrade is finished, install VirtualBox-3.2.

If you are running SELinux you may be locked out of your machine and required to boot to single user mode to fix up your machine. Redhat bug 702865 describes a fix as:

1. setenforce 0
2. yum remove selinux-policy selinux-policy-targeted
3. rm -rf /etc/selinux/targeted

If, after upgrading you want selinux back:

1. yum install selinux-policy selinux-policy-targeted
2. fixfiles restore
3. reboot

Fedora 13

First install the new fedora 13 gpg key. You may wish to verify this package against https://fedoraproject.org/keys and the fedora ssl certificate.

rpm --import https://fedoraproject.org/static/E8E40FDE.txt

Upgrade all packages with

yum update yum
yum clean all
yum --releasever=13 distro-sync

• Testing shows that upgrading within X will cause X to hang when installing bitmap-fonts-compat, but it can be done anyway by making sure that bitmap-fonts-compat is the last package to update ... YMMV.
• Intel video no longer works without Kernel Mode Setting (KMS). You must remove 'nomodeset' from your /etc/grub.conf kernel command line, if present.
• nVidia video no longer works without kernel mode Setting (KMS). You must remove 'nomodeset' from your /etc/grub.conf kernel command line, if present.
• It may be necessary to run "db_recover -h /var/lib/ldap/ ; chown ldap:ldap /var/lib/ldap/*" once in order for slapd to start.

(Legacy) Instructions to upgrade using yum

1. Back up your system

Back up any personal data to an external hard drive or to another machine. If there is some unrecoverable error that requires a fresh install, you don't want to lose any data.

2. Read about common problems

Further down in this page there is a list of common problems for specific versions. Some of them require attention before the upgrade.

General advice on upgrading Fedora can be found on the Upgrading page. You should also read the Installation Guide and Release Notes for the version you plan to upgrade to - they contain important information regarding upgrading issues. Finally, check the list of Common F41 bugs.

3. Clean Stuff

Review and remove all .rpmsave and .rpmnew files before and after upgrading. (And if you have selinux enabled then remember to check security context if you move config files around.)

Find unused config files
Merge and resolve the changes found by the following script: for a in $(find /etc /var -name '*.rpm?*'); do diff -u $a ${a%.rpm?*}; done

Now is a good time to remove packages you don't use - especially non-standard packages.

Find and review "unused" packages
You can find packages not required by other packages with the tool package-cleanup from the yum-utils package: yum install yum-utils; package-cleanup --leaves. These packages could be candidates for removal, but check to see whether you use them directly or if they are used by applications not backed by rpm packages. Remove them with yum remove package-name-and-version.
Another useful tool for cleaning up unused packages is rpmreaper. It's an ncurses application that lets you view rpm dependency graph and mark packages for deletion. Marking one package can make other packages leaf, which you can see immediately, so you don't have to run the tool several times to get rid of whole sub-tree of unused packages. Install with: yum install rpmreaper.

Find and review "lost" packages
You can find orphaned packages (ie packages not in the repositories anymore) with: package-cleanup --orphans. This will also show packages which have been partially uninstalled but where the "%postun" script failed.

Then remove all traces of the version you are leaving from the yum cache in /var/cache/yum.

yum clean all

4. Switch repositories

Update the packages that decide which yum repositories to use.

rpm -Uhv http://archives.fedoraproject.org/pub/archive/fedora/linux/releases/<ReleaseNumber>/Fedora/i386/os/Packages/fedora-release-*.noarch.rpm

Note: Some users have reported that attempting to download through rpm returns a 550 response and "transfer failed" messages. If this is the case, you can download the rpms and install locally. Consider:

wget ftp://dl.fedoraproject.org/pub/archive/fedora/linux/releases/<ReleaseNumber>/Fedora/<Arch>/os/Packages/fedora-release-*.noarch.rpm
rpm -Uvh fedora-release-*.noarch.rpm

Make sure the new repo files isn't placed as .rpmnew files, perhaps by

mv /etc/yum.repos.d/fedora-updates.repo.rpmnew /etc/yum.repos.d/fedora-updates.repo
mv /etc/yum.repos.d/fedora.repo.rpmnew /etc/yum.repos.d/fedora.repo

If you have 3rd party repositories configured, you may need to adjust them for the new Fedora version. If you switch from one Fedora release to another there is often nothing that needs to be done. If you switch to Rawhide from a standard Fedora release (or vice versa) then most of the time you will need to install the Rawhide release RPMs from the 3rd party repository as well (or the standard ones, if switching back).

Note that the upgrade is likely to fail if there are outdated dependencies from packages not backed by a yum repository or backed by a repository which isn't ready for the new version.

5. Do the upgrade

If using selinux make sure it is in permissive mode - neither disabled nor enforcing.

It is a good idea to do the upgrade outside the graphical environment. Log out of your graphical desktop and then go to a text console

ctrl + alt + F2

log in as root, and go into runlevel 3

telinit 3

Once a live upgrade is started do not stop the upgrade by rebooting, killing the process, or by any other method until it is complete, doing so will cause the affected system to be in a mixed state. Partially the old release and partially the new release. In this state the system will not be reliable and will not operate as expected. The only way to recover from this is to reinstall if it occurs.

Update all packages at once
Although you can try to update the update machinery first with yum update rpm\* yum\* or with glibc, in both cases dependencies might expand it to an almost full upgrade anyway, and simply doing a full upgrade might be more safe (but your mileage may vary). (In particular, do not upgrade rpm and yum separately if upgrading to a system with a different version of python e.g. 2.4 to 2.5.)

Upgrade all packages with

yum upgrade

yum might complain about conflicts or requirements. That is probably because you have used non-standard repositories or installed non-standard packages manually. Try to guess which packages cause the problem (or at least is a part of the dependency chain) - uninstall them and try again. Remember to install the packages again if they are essential.

Ensure that all (new) essential packages from the new version are installed with

yum groupupdate Base

You might want to update other groups too, see

yum grouplist

For example

yum groupupdate "GNOME Desktop Environment" \
"Development Tools" "Server Configuration Tools" \
"Hardware Support" "Sound and Video" \
"Graphical Internet" "Fonts" \
"Games and Entertainment" "Printing Software" \
"Administration Tools" "Office/Productivity" "System Tools"

6. Preparing for reboot

Before booting you should usually install the bootloader from your new grub by running

/sbin/grub-install BOOTDEVICE

- where BOOTDEVICE is usually /dev/sda (If you get an error '/dev/sda does not have any corresponding BIOS drive' from that, then try /sbin/grub-install --recheck /dev/sda.)

Also, the order of init scripts could have changed from the previous version. A command to reset the order is:

cd /etc/rc.d/init.d; for f in *; do /sbin/chkconfig $f resetpriorities; done

Again, run package-cleanup --orphans to find packages that haven't been upgraded.

Fedora 12

• All Fedora 12 RPM packages use XZ/LZMA compression, which is only supported by the rpm-4.7.1-1 update (or later) for Fedora 11. Before upgrading from Fedora 11, run yum update rpm.
• It is not possible to update directly from Fedora 10 or earlier with yum. To upgrade from Fedora 10, first update to Fedora 11, and then upgrade from Fedora 11 to Fedora 12. If you are using a release before Fedora 10, use one of the official upgrade methods or backup your data, perform a fresh installation, and restore from backup.
• Upgrades from Fedora 11 might currently fail with missing dependencies (for example related to totem-gstreamer or PolicyKit-kde). The dependencies are ok, but Fedora 11 yum has problems resolving them (bug 519172).
• If your root filesystem does not mount with 'defaults' options, you will need to add the root filesystem's mount options to grub.conf with the kernel flag rootflags - initramfs does not pick up the options the way initrd used to. e.g. if you mount your root with the flags data=journal,relatime then you would add to the end of the 'kernel' line in grub.conf: rootflags=data=journal,relatime . After updating to dracut-004-4.fc12 or later, this kernel flag can be removed.

Fedora 11

• It is only possible to upgrade to Fedora 11 from an updated Fedora 10. Older systems must be upgraded to Fedora 10 first. The RPM format has been changed for Fedora 11, so the older rpmlib must be updated first. Otherwise the process will fail with rpmlib(FileDigests) dependency problems. Support for the new format has been backported to Fedora 10 in an update.

yum update rpm

• New initrd built when installing a new kernel while running Fedora 10 might fail. To solve that boot with an old kernel (to get the new userspace) and (re)install the new kernel.
• Systems with PAE support (indicated by pae in /proc/cpuinfo) should use kernel-PAE.i686. The new kernel must be changed/installed manually: Set DEFAULTKERNEL=kernel-PAE in /etc/sysconfig/kernel and yum install kernel-PAE. Refer to Dave Jones' blog post for details.
• The yum update step should NOT be run inside a gnome desktop session/gnome-terminal. 494046 could result in a unusable install when gnome-terminal segfaults during the upgrade. Update should be run in a vty, runlevel 3, or a screen session.
• fedora-release-11-1.noarch changes the yum mirrorlist URL so that it uses a "metalink", but the version of yum currently in F10 doesn't understand this syntax, leading to yum downloads failing with this error message:

YumRepo Error: All mirror URLs are not using ftp, http[s] or file.
 Eg. </metalink>/

This is 498720. Workaround is to manually edit the URL in /etc/yum.repos.d/fedora.repo as described at https://www.redhat.com/archives/fedora-list/2009-June/msg00783.html

• Some packages in Fedora 10 are regarded as newer than those supplied by Fedora 11 and its updates repository. These include ntpd, ntpdate (506040, 504980), unique, unique-devel, eclipse-changelog, eclipse-svnkit and svnkit. You may wish to remove these before performing the upgrade, then reinstall them afterwards. Doing so may require --nodeps.
• Some i386 packages in Fedora 10 are replaced with i586, i686 or x86_64 packages in Fedora 11. These include gpm.i386, glibc-2.9-3.i386. You may wish to remove these before performing the upgrade, then reinstall them afterwards. Doing so may require --nodeps.
• mplayer-1.0-0.104.20090204svn.fc10 from the RPM Fusion repository has a dependency on libfaad.so.0 that the depsolve doesn't find, but rpm_check_debug does. You may wish to remove mplayer before performing the upgrade, then reinstall them afterwards. Doing so may require --nodeps.

Fedora 10

• Check the Fedora 10 Installation Guide and Fedora 10 Release Notes
• fedora-release packages can be found at rpm -Uvh ftp://dl.fedoraproject.org/pub/fedora/linux/releases/10/Fedora/i386/os/Packages/fedora-release-*.noarch.rpm
• GDM starts on tty1 by default, not tty7.
• anaconda by default now refers to encrypted devices by UUID, not by device name, to better handle device name changes. An encrypted device using partition /dev/sda1 with UUID 12345678-1234-1234-1234-123456789012 would now be called /dev/mapper/luks-12345678-1234-1234-1234-123456789012 instead of /dev/mapper/luks-sda1. You may wish to make the same changes on your system on upgrade; if you do, make sure to run mkinitrd afterwards.
• If you use non-root encrypted devices, you will need to re-run mkinitrd for any older kernels that you want to still boot on the system.
• It's been reported that there are problems with Pidgin i.e.:
  • pidgin-2.5.2-2.fc9.i386 from installed has depsolving problems - Error: Missing Dependency: libedataserver-1.2.so.9 is needed by package pidgin-2.5.2-2.fc9.i386 (installed)
  • the fix was to remove Pidgin

Fedora 9

• Check the Installer section in Release Notes)
• After "the incident" you should use updated fedora-release packages from rpm -Uvh ftp://dl.fedoraproject.org/pub/fedora/linux/updates/9/i386.newkey/fedora-release-*.noarch.rpm
• If you are running as a xen Host (Dom0) you should not upgrade to Fedora 9. [1]
• Upgrading the thunderbird package will fail . The current workaround is yum remove thunderbird; yum install thunderbird>.
• If you are upgrading to Fedora 9 and use emacs, you must upgrade to the latest version of emacs for your prior release to ensure a clean upgrade. Fedora 8 users must have emacs-22.1-10.fc8 or later, while Fedora 7 users must have emacs-22.1-7.fc7. Bug report here.
• Due to the switch from sysvinit to Upstart , it is recommended that users who do an upgrade on a live filesystem to Fedora 9 reboot soon afterwards. If you do not reboot, you may want to take careful note of any '/etc/inittab.rpmsave' or '/etc/inittab.rpmnew' files. When rebooting after an upgrade then be aware that sysvinit that was used to start the system can't be used to shut it down.
• You may find that after upgrading, you can no longer log in to GNOME. If this is the case, use ctrl-alt-F1 to access command prompt, then check the tail of .xsession-errors. If you see a comment like "(named color or font does not exist)", then you may be able to fix this using http://matthew.peach.net.nz/2008/05/18/how-to-fix-badname-named-color-or-font-does-not-exist-error/.

Fedora 8

• New pulseaudio and flash support packages must be installed manually:

yum groupinstall sound-and-video gnome-desktop (if you are using Gnome)
yum groupinstall sound-and-video kde-desktop (if you are using KDE)
yum install libflashsupport paman padevchooser

• The Fedora 7 avahi packagage fails to uninstall; remove it manually with

rpm -e --noscripts avahi-0.6.17-1.fc7

• On 64 bit machines (x86_64) you may have to remove some 32 bit libraries. For example dbus:

yum erase dbus.i386

• You may find it easier to use the yum shell to stack these commands in a single transaction:

yum shell
> erase dbus.i386
> update
> run

• A lot of KDE packages are no longer multilib in F8. If you are on x86_64, you will need to erase the i386 versions of these packages by hand, like so:

yum erase kde{accessibility{,-devel},base,edu,graphics,multimedia{,-extras},network,sdk,utils{,-devel},webdev}.i386

• Various other packages including Beryl are no longer in F8 and not obsoleted:

yum erase beryl\*

Fedora 7

• Fedora 7 replaces the old IDE subsystem with libata. Drive device names which previously started /dev/hd.. will become /dev/sd.. after the upgrade. /dev/hda1 will usually become /dev/sda1, although there may not be a direct relationship between the old and new device names (for example hdd does not necessarily become sdd). Before you reboot be sure to change all references to /dev/hd.. in your config, especially /etc/fstab - where it however may be simpler to refer to filesystems by label (check out the programs blkid, tune2fs, and mlabel). LVM Volume names are not affected. In /boot/grub/device.map change /dev/hd.. to /dev/sd.. before running grub-install - and don't change (hd0). Changing /boot/grub/grub.conf may also be required.

• The libata layer represents all hard disks as SCSI disks, which are limited to 15 partitions in the kernel. IDE hard disks with more than 15 partitions are not supported in Fedora 7.

• On a system which has been upgraded from releases prior to FC6 you may need to remove up2date and rhnlib rpm -e rhnlib up2date

• If you see the message package gpm-1.20.1-84.fc6 (which is newer than gpm-1.20.1-83.fc7) is already installed when performing a yum update uninstall and reinstall the gpm package.

• If you had installed Suns jre it might be removed during upgrade. Be aware that jre-6u1/jre-1.6.0_01-fcs requires compat-libstdc++-33.

• If you get Error: Missing Dependency: python(abi) = 2.4 is needed by package libxml2-python and/or Error: Missing Dependency: python(abi) = 2.4 is needed by package rpm-python during your yum upgrade ensure you have done a yum clean all and check that both the base and updates repositories are available from your mirror. You will need the updates repo for the upgrade to succeed.

• Since Fedora Core 6 uses python 2.4 and Fedora 7 uses python 2.5, do not try and run yum upgrade yum\* rpm\* before upgrading everything else - this succeeds but leaves yum in an unusable state.

Fedora Core 6

• After upgrading, you cannot use the DVD ISO loopback mounted as a repository directly (e.g. no baseurl=file:///mnt/fc6/) because the repodata files contain URLs of type media:// which yum can't handle. You can use the loopback-mounted DVD for the initial upgrade from FC5 -> FC6, just not thereafter. After removing the baseurl=file:/// option from your fedora-core.repo file, remember to yum clean all.

Fedora Core 4 -> Fedora Core 5

• Before upgrading update your kernel first.

• Make sure you have the latest kernel ABOVE 2.6.14 and uninstall all kernel versions before it, otherwise initscripts and a few other packages will conflict and prevent the ugprade. Use something like the command below

• Install fedora-release in the Fedora Core 5 version. Check our mirror sites for a faster download.

• Update your system

• After upgrading, selinux permissions may be incorrect, causing interesting failures such as metacity not starting. As suggested below, to fix this, run touch /.autorelabel and reboot

Fedora Core 3 -> Fedora Core 4

• make sure you're on the latest version of yum for FC3:

yum update yum

• Download fedora-release for FC4:
• wget http://yourmirrorhere/pub/fedora/linux/core/4/$yourarch/os/Fedora/RPMS/fedora-release-4-2.noarch.rpm
• rpm -Uvh /path/to/fedora-release-4-2.noarch.rpm
• yum update ( This will start the upgradation process to FC4)
• Once you have updated, run: yum groupinstall "GNOME Desktop Environment" (change GNOME to KDE if appropriate). That'll get you some new packages that were added in FC4 like Evince and NetworkManager.
• Some new groups were added - yum grouplist - check out Eclipse and Java Development in particular.

Problems:

• Postgres database in FC4 has a newer incompatible format from the previous versions. Manual dump and restoration is required if you using this database. Look at the postgres documentation for more details on this.

• kernel requires kernel-utils (or vice-versa) run:

yum update kernel
reboot
yum remove kernel-2.6.11\*FC3\*
yum upgrade
reboot

• Takes forever and a day and hangs at:

Reading repository metadata in from local files This means you have too much crap in your rpmdb and it's taking a while to go through it. Check out how many kernels you have installed: rpm -q kernel kernel-smp and remove the old ones you may also try running: rpm --rebuilddb and see if it helps any. I've found that removing old kernels and rebuilding the rpmdb helped this problem.

• x86_64 upgrade requires removal of certain i386 packages before update

yum remove perl.i386
TODO: Add list of bonobo i386 packages to remove here

• Error: Missing Dependency: libpython2.3.so.1.0 is needed by package koffice

koffice was removed from Fedora Core and not moved to extras - so there's no way to complete an upgrade to FC4 without removing it. run:

yum remove koffice

then your update should complete.

• x86_64 systems seems to hang during "Running Transaction Test"

Seems related to bug 155730 -- Maybe a update to the "rpm-4.3.3-3.0.fc3" (currently in updates-testing) might solve this. Ugly workaround (to be done as root):

: > /var/log/lastlog

• GPG check fails with this message :

warning: rpmts_HdrFromFdno: V3 DSA signature: NOKEY, key ID db42a60e
public key not available for <some_package>
Retrieving GPG key from file:///etc/pki/rpm-gpg/RPM-GPG-KEY-fedora

The GPG key at file:///etc/pki/rpm-gpg/RPM-GPG-KEY-fedora (0x4F2A6FD2)
is already installed but is not the correct key for this package.
Check that this is the correct key for the "Fedora Core 4 - i386 - Base" repository.

Edit /etc/yum.repos.d/fedora.repo and replace the line

gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY-fedora

with:

gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY

• SELinux gets messed up. I had to redownload the selinux-policy-targeted RPM from my mirror, do a forced uninstall of the old package, clean out the .rpmnew and .rpmsave files then install the new policy. Then I had to run "/sbin/fixfiles relabel" and wait. If I didn't do that, I found I couldn't log in.
• If SELinux does give you problems, it can be disabled on bootup by pressing "a" on the grub bootup menu, then adding "enforcing=off" to the arguments list.

• RPM complains about missing diskspace, but there is a lot free. If you are really sure your diskspace is enough, set

diskspacecheck=0
</code> in <code>/etc/yum.conf</code>.
The RPM transaction will need a lot of diskspace temporarily (much more than one might think), so be careful.
Don't forget to remove that later after the upgrade.

* Yum complains about files conflicting in the GCC package.  This system was upgraded from Fedora Core 1.  Remove the gcc32 package:
<pre>
sudo rpm -e gcc32

• Some of your modules don't load after the upgrade. When you investigate modprobe doesn't work as expected. Run depmod:

sudo depmod -ae -F /boot/System.map-[your <code>uname -a</code> kernel version here] 

• Yum has been reported to be hanging after upgrading. Updating the pysqlite package might fix it.

Fedora Core 2 -> Fedora Core 3

Go here for most of this info: https://web.archive.org/web/20080430083623/http://linux.duke.edu/~skvidal/misc/fc2-fc3-update-with-yum.txt

Fedora Core 1 -> Fedora Core 2

Just Upgrade using anaconda - save yourself a world of pain. But if you really want to try, there's these hints by Seth Vidal: https://web.archive.org/web/20080720005856/http://linux.duke.edu/~skvidal/misc/fc1-fc2-yum-hints.txt