How do I migrate my Bright CentOS 8 cluster to Rocky Linux 8 ?

Bright 9.1, 9.0 and 8.2 Clusters running CentOS 8 can be migrated to Rocky Linux 8. This involves migrating the head nodes, software images, and node-installer images (Bright 9.1 and 9.0).

Prerequisites

Support for rocky requires Bright Cluster Manager and associated packages to be at or above 9.1-9 released after Oct 12th, 2021, 9.0-17 released after Nov 3rd 2021 or 8.2-27 released after Dec 13th 2021. Please make sure your systems are online or have access to a cached copy of the Bright updates repository at the latest versions.

Rocky Linux migration script

Download the Rocky Linux migration script to the active head node, and passive head node (in a HA setup).

$ wget https://raw.githubusercontent.com/rocky-linux/rocky-tools/main/migrate2rocky/migrate2rocky.sh -O /root/migrate2rocky.sh

$ chmod u+x /root/migrate2rocky.sh

Migrate head nodes

1. Run the migration script on the active head node, and passive head node (in a HA setup).

$ /root/migrate2rocky.sh -r

2. Before rebooting the head node, re-create the grub config (must be done on both head nodes in a HA setup).

$ grub2-mkconfig -o /boot/grub2/grub.cfg

3. Reboot head nodes.

Migrate software images

1. Migrate the software images on the active head node. The following must be done for every CentOS 8 software image, in case there are multiple software images. It is possible you may receive some error messages while running this step of the migration due to running inside the chroot’ed environment, it is safe to ignore these errors (eg, ” System has not been booted with systemd as init system” and/or “dracut: no ‘/dev/log’ or ‘logger'”)

$ cp /root/migrate2rocky.sh /cm/images/<software image>/root

$ cm-chroot-sw-img /cm/images/<software image> /root/migrate2rocky.sh -r

Note: If you get an error similar to Can't find EFI mount. No EFI boot detected. please use the following alternative syntax.

$ systemd-nspawn -D /cm/images/<software image> -- /root/migrate2rocky.sh -r

Note: Ignore reboot message from migration script.

Note: For software images that are placed in locations other than /cm/images, the path to the software image must be updated accordingly in the commands above.

2. Sync new software images to other provisioning nodes.

$ cmsh
  softwareimage 
  updateprovisioners

Migrate node-installer image

Important: This step is only applicable for Bright clusters running version 9.0 and above.

Important: In some cases dnf-plugins-core might not be installed in the node-installer image. It must be installed as follows before proceeding with the migration.

$ cm-chroot-sw-img /cm/node-installer dnf install dnf-plugins-core

1. Migrate node-installer image on active head node.

$ cp /root/migrate2rocky.sh /cm/node-installer/root

$ cm-chroot-sw-img /cm/node-installer /root/migrate2rocky.sh -r

Note: If you get an error similar to Can't find EFI mount. No EFI boot detected. please use the following alternative syntax.

$ systemd-nspawn -D /cm/node-installer -- /root/migrate2rocky.sh -r

Note: Ignore reboot message from migration script.

Note: In a multi-os setup with a CentOS 8 head node, the node-installer image will be located in /cm/node-installer-centos8-x86_64. The path must be adjusted accordingly in the above commands.

2. Trigger update of node-installer FSPart

$ cmsh 
  fspart
  trigger -t node-installer

Reboot nodes

1. Reboot regular nodes that have been configured to use the migrated software image.
2. Terminate and start cloud compute nodes to use the latest node-installer image that has Rocky Linux support.