8.6 KiB
migrate2rocky -- Conversion Script
Running this script will convert an existing CentOS 8 system to Rocky Linux 8.
Usage
./migrate2rocky.sh -h
├── -d # --> Do not update before conversion
├── -h # --> Display this help
├── -o # --> Work in offline mode
├── -r # --> Convert to Rocky
└── -V # --> Verify switch
[!! USE WITH CAUTION !!]
Disk Space Requirements
Please note the following disk space requirements. These requirements may vary from one system to another. Offline mode may require further space for rpm packages and repository metadata. Failure to have adequate disk space available may result in migrate2rocky leaving the system in an unstable state:
/usr 250M
/var 1.5G
/boot 50M
Recommended Practice
When running this script, especially via a remote session, it is highly recommended to enter a screen or tmux session before running. If a standard ssh or terminal session, such as the Cockpit Terminal window, is disrupted, the script will die and leave the system in a potential unrecoverable state. For more on tmux sessions, please see: https://github.com/tmux/tmux/wiki
Known Issues
Running the script in Cockpit's Terminal Screen will be interrupted
Do not run this script through the Terminal screen built into Cockpit. As the script runs the upgrades, Cockpit will be restarted and Terminal connection will disconnect, thus stopping the script and leaving the system in an unrecoverable state. It may be possible to launch a screen or tmux session from the Cockpit Terminal, but USE AT YOUR OWN RISK.
EL8.0 Migrations
If you are attempting to migrate a system that has not been updated since 8.0
then you must run dnf update
before attempting the migration.
If you are migrating from CentOS 8.0 then you must manually fix the baseurls of
the CentOS repositories before running dnf update
:
sed -i -r \
-e 's!^mirrorlist=!#mirrorlist=!' \
-e 's!^#?baseurl=http://(mirror|vault).centos.org/\$contentdir/\$releasever/!baseurl=https://dl.rockylinux.org/vault/centos/8.5.2111/!i' \
/etc/yum.repos.d/CentOS-*.repo
Migration in offline mode
Offline mode has received minimal testing and only for the CentOS 8.5 -> Rocky 8.5/8.6 migration from a minimal (straight from installation or fully updated) system using CentOS 8.5 DVD and Rocky 8.5/8.6 DVD ISOs as offline repositories, mounted respectively under /mnt/centos (hardcoded into and forced by the script only for the CentOS case - for other EL8 distros you will need to manually define local repo paths under /etc/yum.repos.d/ before launching the script) and /mnt/rocky (hardcoded into and always forced by the script - please note that you will also need the Rocky GPG key in /mnt/RPM-GPG-KEY-rockyofficial) .
Please make sure that you do not inadvertently swap the current-EL8/Rocky local repositories when mounting ISOs.
Package downgrades may happen, depending on the update status of the running EL8 distro and on the content of the local repositories (e.g. using a Rocky 8.5 DVD ISO for offline migrating a CentOS 8.5 with all latest-before-EOL updates applied). Note that if you are disabling pre-migration updates then you will need only the base repositories' metadata (repodata subdir) for the currently running EL8 (can be copied from original installation ISO).
If you installed packages from further repos of the running EL8 distro (devel, extras, ha, powertools) then you will need those same repos enabled and their Rocky equivalents locally available (under /mnt/rocky/{Extras,HA,PowerTools,Devel}) before starting the migration, otherwise make sure that those additional repos are disabled.
Migration with pre-updating disabled
Disabling system updates before migration has been introduced and tested only as a convenience option for the offline migration mode but it should be carefully tested on a non-production system before attempting it.
Custom replacements of default repositories
This script expects the original repository configuration being present, as
well as enabled (i.e. for CentOS the baseos
repo configuration in the
/etc/yum.repos.d/CentOS-Linux-BaseOS.repo
file has to be present and enabled).
Offline mode will require at least the repository metadata (repodata subdir) for
the baseos
and appstream
repos to be available (for CentOS their path is
hardcoded as /mnt/centos/{BaseOS,AppStream} and can be copied or mounted from a
CentOS ISO).
Also make sure that there are no other repositories which could interfere
with the original configuration.
Any distribution that has had its core repositories altered, removed, duplicated or overridden may cause migrate2rocky to break or corrupt the system when run. Any attempt to migrate such systems, even after reversing the changes made by such software, is not supported in any way. In all cases you should backup your system before using migrate2rocky and USE AT YOUR OWN RISK.
This especially happens on systems configured with a centralized package management like Katello (RedHat Satellite 6) or Uyuni (RedHat Satellite 5, SUSE Manager).
RHEL migrations show error messages during conversion
Installing : rocky-release-8.3-13.el8.noarch 2/5Error unpacking rpm package rocky-release-8.3-13.el8.noarch
...
error: unpacking of archive failed on file /usr/share/redhat-release: cpio: File from package already exists as a directory in system
error: rocky-release-8.3-13.el8.noarch: install failed
...
Error: Transaction failed
This results from conflicts in the directory structure of RHEL with that of RockyLinux. migrate2rocky will detect the issue and go on to remove the conflicting directory and install rocky-release with the rpm command.
No matches found for the following enable plugin patterns: config-manager
The above error message is a display bug in dnf. It does not affect the actual dnf command or the migration. You may safely ignore this message. (RHBZ#1980712)
Grub still shows kernel entries from previous installation
This is normal. The running kernel cannot be safely removed when migrate2rocky is run. The RockyLinux kernel should come up as the default highlighted kernel on reboot but the other ones will remain until they are removed or replaced by newer kernels. If you want you can manually remove the old kernels after reboot with dnf or rpm.
Symbolic links to Java programs in /etc/alternatives
are gone
After migrating a system with an OpenJDK package installed you might encounter
that java does not work any more. There is a bug in the OpenJDK packages that
leads to losing the symbolic links in /etc/alternatives
during the migration.
A bug report against the upstream packages has been filed
here. As a workaround
you can use the following script to recreate the package default alternatives:
rpm -qa --scripts java-{1.8.0,11}-openjdk-{headless,devel} | sed -n '/postinstall/, /exit/{ /postinstall/! { /exit/ ! p} }' | sh
IPA fails to start after migration
This issue is caused by a version mismatch due to the way that modules work that trick ipa into thinking that the package was downgraded even if it was not. To fix this issue run the following command after migration:
ipa-server-upgrade --skip-version-check
Note: Since ipa-server-upgrade is a java program you will likely have to run the command to mitigate the "Symbolic links to Java programs..." issue above before running this command.
CentOS SIG repositories disappear after migrating to RockyLinux.
This is because the centos-release-* packages that contain the .repo files for the individual repositories depend on centos-release. Storage sig and related release packages should be available soon from RockyLinux. In the meantine you can use a command like the following to install the .repo files and continue to use the repository from CentOS (note please substitute the URL to the release package for the repo that you need):
rpm2cpio <(curl http://mirror.centos.org/centos/8/extras/x86_64/os/Packages/centos-release-gluster9-1.0-1.el8.noarch.rpm) | cpio -iD/ \*.repo
Latest Version
The latest version of this script can be found here.
Debugging
The migrate2rocky
script pipes everything shown on stdout
and stderr
to
/var/log/migrate2rocky.log
.
If you run in to issues executing this script, please submit an issue here.
Make sure to include the output log, and remove any sensitive information. (if any)
Feel free to create a pull request if you think you've got the fix.