Search Docs...
⌘ K
Prerequisite
Check cluster resources to ensure both computing and storage resources support single-node failure scenarios.
Run the command uname -a to check and record the kernel version before the upgrade.
Ensure that the cluster completes the upgrade; otherwise, the kernel may become incompatible with internal services of ACOS.
Follow the Entering maintenance mode for a self-check and ensure that all check items meet the requirements.
If virtual machines with vGPUs installed exist on the host to be upgraded, you need to shut these virtual machines down.
Run the command zbs-tool service list to identify the Meta Leader node in the cluster.
Upload the upgrade ISO file of the current version to the Meta Leader node, then run the following command to prepare the repository required for the kernel upgrade. After that, record the storage IP of the node. In the command below, <iso_path> represents the absolute path of the ISO file on the node.
cluster-upgrade prepare_repo <iso_path>
After execution, if the output displays prepare repo success, it indicates the repository is ready.
Log in to AOC, select the host on which this node resides from the list of hosts in the cluster, click the ellipsis (...) on the right and choose Entering maintenance mode.
Run the following command to prepare the repository, where <repo_ip> refers to the storage IP of the Meta Leader node.
cluster-upgrade config_repo --repo_ip <repo_ip>
Once the command execution finishes, if the output message shows config repo success., it indicates the repository required for the node kernel upgrade has been configured successfully.
Execute the following command to upgrade the kernel. When the message Upgrade done appears, it indicates a successful upgrade.
skc upgrade
Log in to AOC, select the target host from the cluster's host list, click on the ... on the right side, and choose Reboot. After that, enter the reason for rebooting and then restart the host.
Run the uname -a command to check if the kernel has successfully upgraded to the target version. If the upgrade does not meet expectations, refer to the section Rolling back the kernel version to roll back the kernel.
On the overview page of the host in AOC, view the duration the current host has been in maintenance mode. Click on Exit maintenance mode on the right side of the prompt.
The system will open a check dialog box to perform exit checks. After the check is complete, the dialog box will show the results of all check items.
Switch to other non-Meta Leader nodes and perform the above steps until all non-Meta Leader nodes have completed the upgrade.
Run the zbs-tool service list command to confirm whether the Meta Leader has switched to another node. If the Meta Leader has not switched to another node, run the systemctl stop zbs-metad command to trigger the Meta Leader switch.
Run the zbs-meta chunk list command to verify that the status of each Chunk in the current cluster is CONNECTED_HEALTHY.
Log in to AOC, select the host on which this node resides from the list of hosts in the cluster, click the ellipsis (...) on the right and choose Entering maintenance mode.
Run the following command to prepare the repository needed for the kernel upgrade, where <repo_ip> refers to the storage IP of the node.
cluster-upgrade config_repo --repo_ip <repo_ip>
After executing the command, if the output message shows config repo success., it indicates that the required repository is configured successfully.
Execute the following command to upgrade the kernel. When the message Upgrade done appears, it indicates the upgrade is successful.
skc upgrade
Log in to AOC, select the target host from the cluster's host list, click on the ... on the right side, and choose Reboot. After that, enter the reason for rebooting and then restart the host.
Run the uname -a command to check if the kernel has successfully upgraded to the target version. If the upgrade does not meet expectations, refer to the section Rolling back the kernel version to roll back the kernel.
On the overview page of the host in AOC, view the duration the current host has been in maintenance mode. Click on Exit maintenance mode on the right side of the prompt.
The system will open a check dialog box to perform exit checks. After the check is complete, the dialog box will show the results of all check items.
After completing the upgrade on all nodes, confirm the cluster status is normal through AOC.
Due to limited storage space in the /boot partition, repeated kernel upgrades may fill up this storage space. In this case, execute the skc clean command to clean up unused drivers and kernel packages.