Common bugs and issues

Below you will find the most common bugs and issues, that we encounter at Bugzilla, and their solutions.

Bug report issues

These issues require more information from the reporter.

Too old version of Anaconda

Issue:

The bug is reported against a too old version of the operating system. It it possible that the code has changed and the problem no longer exists.

Solution:

Are you able to reproduce the problem with Fedora XY?

Missing logs

Issue:

There are no useful logs attached to the bug.

Solution:

Please, attach all files with installation logs, especially the file named syslog. You can find them during the installation in /tmp or on the installed system in /var/log/anaconda/.

If you are gathering the log files manually - outside of the automatic bug reporter - please make sure that either the file syslog or journal.log is included. If you are in the Live CD environment, and neither file exists, please create the journal dump manually by running journalctl > journal.log and attach the file.

DBus issues

Anaconda runs several DBus modules and communicates with them from the user interface, so you can easily come across a DBus-related issue.

Traceback of DBusError

Issue:

Anaconda fails with the dasbus.error.DBusError exception. This usually happens when a DBus module raises an unexpected exception. Anaconda shows a traceback only for the DBus call, so it is necessary to look up a traceback of the DBus module to have complete information about the bug.

Solution:

You can find the original exception in the logs (usually in syslog or in the output of journalctl).

Example:

rhbz#1828614

Installation environment issues

You can find here issues related to the installation environment. Anaconda usually runs in the stage2 environment provided by boot.iso, in Live OS, in a mock environment or locally.

Mismatched stage2

Issue:

Anaconda fails early in stage2 with an exception “ValueError: new value non-existent xfs filesystem is not valid as a default fs type”.

Solution:

This error occurs when initrd.img, vmlinuz and the repository (or stage2) are not from the same media or location.

Example:

rhbz#1169034

Out of memory

Issue:

Anaconda fails in stage1 with a message “Failed writing body” or “No space left on device” in the dracut logs. This usually happens when installing from http or ftp source on a machine with insufficient memory size. See the minimal requirements for RHEL.

Solution:

Increase the memory size or try installing from NFS, CD-Rom or HDD source.

Example:

rhbz#1630763

Changes in Live OS

Issue:

The Live OS requires changes.

Solution:

Reassigning to spin-kickstarts.

Changes in boot.iso

Issue:

The boot.iso requires changes.

Solution:

Reassigning to lorax.

Icon issues

Issue:

The Anaconda icons in Live OS requires changes.

Solution:

Reassigning to fedora-logos

Example:

rhbz#1699034

Font issues

Issue:

In the Welcome spoke, there are replacement glyphs (rectangles) instead of characters in a name of a language. This usually means that that there is no font for this language installed in the installation environment.

Solution:

Reassigning to lorax or spin-kickstarts.

Example:

rhbz#1530086

Payload issues

These issues are related to the content that is installed on the target system.

Non-fatal POSTIN scriptlet failure

Issue:

The package installation fails with a message “Non-fatal POSTIN scriptlet failure in rpm package”. The failing package has to fix its scriptlet, because all scriptlets MUST exit with the zero exit status.

Solution:

All RPM errors are fatal during the installation (see the bug 1565123). Reassigning.

Example:

rhbz#1588409

Changes in package groups and environments

Issue:

The reporter wants a new package to be installed by default.

Solution:

The package groups and environments are defined by comps (see https://pagure.io/fedora-comps/). Reassigning to distribution.

Example:

rhbz#1787018

Corrupted ISO

Issue:

The package installation fails with a message “Some packages from local repository have incorrect checksum”. This happens when the packages cannot be accessed, because they are located on a corrupted ISO or an unmounted device.

Solution:

The ISO might be corrupted. Please, try to download it again and verify the checksum.

Example:

rhbz#1551311

Issues with live payload

Issue:

The image installed by the live OS payload requires changes.

Solution:

Anaconda doesn’t create the live image. Reassigning to spin-kickstarts.

Issues with OSTree

Issue:

The installation with the OSTree payload fails.

Solution:

It might be related to the OSTree payload. Reassigning to Colin Walters.

Failed to mount the install tree

Issue:

The payload fails to set up and raises the error “Failed to mount the install tree”. This usually happens when Anaconda is unexpectedly terminated and started again. Some of the Anaconda’s mount points stays mounted and that causes the crash.

Example:

rhbz#1562239

System upgrades

Issue:

The system was upgraded, not installed.

Solution:

Anaconda is not doing system upgrades. That is done by dnf-system-upgrade. Reassigning to dnf.

Missing systemd-machine-id-setup on Live

Issue:

The installation from Live media stopped with traceback about [Errno 2] No such file or directory: systemd-machine-id-setup.

Solution:

None known yet. Try checking the hardware, downloading the Live ISO again and then re-run the installation.

This is probably the first program to run from the installed system. The error can be caused by a corrupted media or failing hardware, although other causes are possible too.

Example:

rhbz#1963778

Storage issues

These issues are related to hardware, partitioning and storage configuration.

Bug in blivet

Issue:

The exception starts in blivet or libblockdev.

Solution:

It seems to be an issue in the storage configuration library. Reassigning to blivet.

Example:

rhbz#1827254

Bug in blivet-gui

Issue:

The exception starts in blivet-gui or there is a problem with partitioning and the reporter used Blivet-GUI as the partitioning method.

Solution:

It seems to be an issue in blivet-gui. Reassigning.

Example:

rhbz#1833775

Failing hardware

Issue:

The logs (journal or syslog) are full of kernel messages about I/O errors. For example:

kernel: [sdb] tag#9 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
kernel: [sdb] tag#9 Sense Key : Medium Error [current]
kernel: [sdb] tag#9 Add. Sense: Unrecovered read error - auto reallocate failed
kernel: [sdb] tag#9 CDB: Read(10) 28 00 1d 04 10 00 00 00 08 00
kernel: print_req_error: I/O error, dev sdb, sector 486805504

Solution:

It looks like a hardware failure. Please, check your hardware.

Example:

rhbz#1685047

LVM on disks with inconsistent sector size

Issue:

The storage configuration fails with an error message mentioning “inconsistent sector size”.

Solution:

LVM is now demanding that all disks have consistent sector size, otherwise they can’t be used together. Please adjust your disk selection to use only disks with the consistent sector size.

Example:

rhbz#1754683

Unlocked LUKS

Issue:

The storage configuration fails with a message “luks device not configured”.

Solution:

Anaconda doesn’t support LUKS devices that are unlocked outside the installer. The device has to be unlocked in Anaconda.

Example:

rhbz#2019455

Undetected partitions

Issue:

When the custom partitioning spoke is entered, it raises an exception with a message: “cannot initialize a disk that has partitions”. Anaconda tries to initialize disks that are supposed to be empty, but there are partitions that were not discovered by kernel after boot.

Solution:

Duplicate of the bug 1825067.

Example:

rhbz#1828188

Too little memory for LUKS setup

Issue:

Anaconda crashes with an exception: No such interface “org.freedesktop.DBus.Properties” on object at path /org/fedoraproject/Anaconda/Modules/Storage/Task/ .

Solution:

The installation environment does not have enough memory to run LUKS setup, and its crash resets the Storage module. In logs, the following lines can be found:

• WARNING:blivet:Less than (...) MiB RAM is currently free, LUKS2 format may fail.

• ui.gui.spokes.storage: Partitioning has been applied: ValidationReport(error_messages=[], warning_messages=['The available memory is less than 128 MiB which can be too small for LUKS2 format. It may fail.'])

• Activating service name='org.fedoraproject.Anaconda.Modules.Storage' (present more than once)

Note that the user must have ignored a warning in the GUI.

Workaround:

There are several possible workarounds:

• Use more memory for the machine,

• use --pbkdf* options in kickstart file,

• change LUKS version to LUKS1,

• disable encryption.

Example:

rhbz#1902464

Using ignoredisk on previous LVM installation

Issue:

When starting installation with automatic partitioning and using ignoredisk kickstart command, it raises an exception with a message: “Selected disks vda, vdb contain volume group ‘vg0’ that also uses further unselected disks. You must select or de-select all these disks as a set.”

Solution:

Anaconda won’t touch disks in ignoredisk kickstart command, however, other disks have part of a Volume Group which is also on disk ignored by the ignoredisk command. To resolve this issue the ignored disks have to be erased manually or by %pre section similar to:

vgchange -an
wipefs -a /dev/vda1 /dev/vda

Example:

rhbz#1688478

Bootloader issues

There issues are related to bootloader issues.

Bug in bootloader

Issue:

The exception is raised during a bootloader installation with a message that usually says “failed to write bootloader” or “boot loader install failed”. Look into program.log or storage.log for more information.

Solution:

Could the bootloader team have a look at this bug, please?

GRUB2 does not detect MD raid (level 1) 1.0 superblocks on 4k block devices

Issue:

Installation failed on grub2-mkconfig command, with actual error like: grub2-probe error disk mduuid/4589a761dde10c78a204bcfd705df061 not found. on block device with 4096 bytes sector size.

Solution:

use workaround.

Workaround:

make your EFI partitions as second disk partitions, i.e. nvme0n1p1 is for / RAID, and nvme0n1p2 is partition for the /boot/efi RAID.

Example:

rhbz#1443144

Disable rhgb quiet

Issue:

The reporter doesn’t want the default boot options rhgb quiet to be used.

Solution:

The installer adds the boot options rhgb quiet only if plymouth is installed. In a kickstart file, you can disable these options with the following snippet:

%packages
-plymouth
%end

Invalid environment block

Issue:

The bootloader installation fails with an exception “failed to write boot loader configuration”. You can find the following message in the logs:

/usr/bin/grub2-editenv: error: invalid environment block

Solution:

Duplicate of the bug 1814690.

Example:

rhbz#1823104

‘utf-8’ codec can’t decode byte

Issue:

Installing the boot loader fails with an exception UnicodeDecodeError. Logs contain a message along these lines:

UnicodeDecodeError: ‘utf-8’ codec can’t decode byte 0x?? in position ??: invalid start byte

Actual byte, position, and byte type (start, continuation, ???) vary.

This is caused by efibootmgr which prints raw non-UTF-8 data to output.

Solution:

Duplicate of bug 2148480.

Example:

rhbz#2238691

User interface issues

These issues are related to the text and graphical user interfaces of the installation program.

Allocating size to pyanaconda+ui+gui+MainWindow

Issue:

Anaconda shows a Gtk warning “Allocating size to pyanaconda+ui+gui+MainWindow without calling gtk_widget_get_preferred_width/height(). How does the code know the size to allocate?”

Solution:

This is an issue in the GTK library: See: https://gitlab.gnome.org/GNOME/gtk/issues/658

Example:

rhbz#1619811

Bug in Gtk

Issue:

When Anaconda is started in the graphical mode, some of the Gtk widgets look weird.

Solution:

Reassigning to gtk3.

Weirdly displayed GUI

Issue:

When Anaconda is started in the graphical mode, the whole screen looks weird.

Solution:

It looks like an Xorg or kernel issue. Reassigning to xorg-x11 for further triaging.

Rotated screen

Issue:

The screen is rotated.

Solution:

It seems to be a problem with drivers. Reassigning to kernel.

Contact:

kernel or iio-sensor-proxy

No video output with the MGA G200e graphics card

Issue:

There is no video output with MGA G200e graphic card and a 1920x1080 monitor.

Solution:

Add vga=795 to the boot options before installation. Alternatively it is also possible to select “Troubleshooting” in the installation image boot menu and install using the basic graphics mode. Please note that the installed system will boot into text mode if installed in basic graphics mode.

Example:

rhbz#2000537

Localization issues

These issues are related to the localization support in Anaconda.

Changes in localization data

Issue:

Languages, locales, keyboard layouts or territories are not correct.

Solution:

This content is provided by langtable. Reassigning.

Example:

rhbz#1698984

Network issues

/mnt/sysroot/etc/resolv.conf issue in Fedora 35

Issue:

Due to systemd update Fedora 35 installation tracebacks with dasbus.error.DBusError: [Errno 2] No such file or directory: '/mnt/sysroot/etc/resolv.conf'.

Happens since systemd-249.10-1.fc35, present also in systemd-249.11-1.fc35, systemd-249.12-1.fc35, … ?

Solution:

Reassign to systemd https://bugzilla.redhat.com/show_bug.cgi?id=2074083.

Workaround:

https://bugzilla.redhat.com/show_bug.cgi?id=2074083#c63. Other option could be: install without updates and update after new system boot.

Example:

rhbz#2083411

Missing /etc/resolv.conf for %post scripts

Issue:

In case systemd-resolved package is not installed to target system the /etc/resolv.conf symlink is not created in installed system root during packages installation and therefore name resolution in %post kickstart scripts does not work. This should not be a problem for the installed system where NetworkManager would handle the /etc/resolv.conf file when systemd-resolved is not present.

Solution:

Anaconda should try hard to avoid handling /etc/resolv.conf file on its own so that it does not interfere with services responsible for it - systemd-resolved or NetworkManager. In case systemd-resolved is not installed there needs to be another mechanism provided by the origin of such a choice, for example a %post script copying the resolv.conf to the chroot environment as suggested in the workaround.

Workaround:

Copy the symlink created by systemd-resolved in installer environment to the chroot in a %post --nochroot script.

For example:

%post --nochroot
if [ ! -e /mnt/sysimage/etc/resolf.conf ]; then
  cp -P /etc/resolv.conf /mnt/sysimage/etc/resolv.conf
fi
%end

The resolv.conf file should be removed after the work in %post scripts requiring it is finished, so that NetworkManager can take care of /etc/resolv.conf management when booting into installed system.

For example in a %post script run in chroot:

%post
# do some stuff
rm /etc/resolv.conf
%end

Example:

rhbz#2101527

Kickstart issues

These issues are related to automated installations that use kickstart files.

Automatic installation in Live OS

Issue:

The reporter would like to run a kickstart installation in Live OS. One of these messages is displayed: Kickstart is not supported on Live ISO installs, please use netinstall or standard ISO. This installation will continue interactively. Alternatively, before Fedora 35: Kickstart is not supported on live installs. This installation will continue interactively.

Solution:

Kickstart installations in Live OS are not supported. Please, run the installation with one of the following types of images:

• netinstall ISO (such as the Server edition of Fedora)

• standard ISO

• boot.iso

Example:

rhbz#1027160

Invalid partitioning in the output kickstart file

Issue:

The kickstart file generated by Anaconda at the end of the installation defines an invalid partitioning.

Solution:

This part of the kickstart file is generated by the storage configuration library. Reassigning to blivet.

Example:

rhbz#1851230

The ignoredisk --only-use command hides installation sources

Issue:

The installer fails to find an installation media on the USB drive if the ignoredisk –only-use= command is specified in a kickstart file.

Workaround:

You can use the harddrive command instead of the cdrom command. For example:

harddrive –partition=sda –dir=/

where sda is the name of the USB device, or use LABEL:

harddrive –partition=LABEL=CentOS-8-3-2011-x86_64-dvd –dir=/

Example:

rhbz#1945779

Missing options of the repo command

Issue:

The repo kickstart command doesn’t support the requested configuration options.

Workaround:

We get a lot of feature requests for the repo command, but we don’t really want to support every repo configuration option. Please, use a repo file to configure the repo.

For example:

# Enable the custom repo.
repo --name "my-custom-repo"

%pre
# Generate the custom repo file.
cat >> /etc/anaconda.repos.d/custom.repo << EOF

[my-custom-repo]
name=My Custom Repository
baseurl=http://my/custom/repo/url/
priority=10
module_hotfixes=1

EOF
%end

The autopart --nohome command doesn’t fill available space

Issue:

The partitioning layout generated by the autopart --nohome kickstart command doesn’t have to fill all available space on the selected disks unlike the autopart command. The default partitioning (used by autopart) is defined in the Anaconda configuration files, for example like this:

default_partitioning =
  /     (min 1 GiB, max 70 GiB)
  /home (min 500 MiB, free 50 GiB)

It means that the size of the / partition can be at most 70 GiB, but the size of /home is unlimited. The --nohome option just skips the request for the /home partition without any other changes. It is the same as running autopart with the following definition:

default_partitioning =
  /     (min 1 GiB, max 70 GiB)

Since the / partition has a limited size, it cannot grow more than that. This behaviour cannot be changed without breaking backward compatibility. Please, use partitioning commands (like part --grow) to create this kind of partitioning.

Solution:

Close as NOTABUG.

Example:

rhbz#1832570