CSI Storage Plugin Upgrade
Before carrying out CSI updates, please carefully read the precautions below. If you have any questions, please consult our technical support.
1. Precautions
- When performing storage plug-in upgrades on the cluster, it is necessary to operate K8s key components. Please operate during the low tide of the business and avoid service launches.
- Do not upgrade by modifying the image of CSI on your own, otherwise, CSI will not work. Please be sure to complete the upgrade of CSI on the console.
- If the version of the cluster is not within our maintenance version, the console will not be able to upgrade directly, see: UK8S Version Maintenance Description.
- 22.09.1: If you want to use RSSD cloud disk, please upgrade CSI to version
22.09.1or above, see details RSSD Cloud Disk Mounting Problem. - 21.09.1: Upgrading from the old version to
21.09.1or above will cause the pod mount points using US3/UFile to fail. If your business uses US3/UFile, please be sure to confirm the current version. If you have any doubts, please contact our technical support.
2. Version Check and Plug-in Upgrade
On the "Plug-in-Storage Plug-in" page of the UK8S cluster console management page, turn on the CSI storage plug-in upgrade function, enable the CSI plug-in function will execute the CSI plug-in query task in the cluster, which takes about 3 minutes, please do not operate the cluster during this process. After the upgrade function is turned on, you can see the version information of the CSI plug-in, and you can upgrade by clicking on "Upgrade CSI".
The upgrade process takes approximately 1 minute, during which the "Current Version" field will be displayed as "Upgrading". After the upgrade is complete, it will display the latest version number. If the upgrade fails, please contact our technical support.
After all nodes have been successfully upgraded, you may close the plug-in upgrade service and reopen it when there is a future upgrade requirement.
3. Manual Upgrade
If the cluster version is not within our maintenance scope, you can manually upgrade the csi, please execute the following commands (you can only upgrade to an old version, not recommended for use with RSSD cloud disks):
UDisk CSI Upgrade
kubectl apply -f https://docs.surfercloud.com/uk8s/yaml/volume/udisk.21.11.2/csi-controller.yml
kubectl apply -f https://docs.surfercloud.com/uk8s/yaml/volume/udisk.21.11.2/csi-node.ymlUFile CSI Upgrade
kubectl apply -f https://docs.surfercloud.com/uk8s/yaml/volume/us3.21.11.2/csi-controller.yml
kubectl apply -f https://docs.surfercloud.com/uk8s/yaml/volume/us3.21.11.2/csi-node.yml4. Change Log
| Version | Update Time | Update Content |
|---|---|---|
| 24.06.07 | 2024.06.07 | Solved the sporadic unloading failure problem when multiple cloud disks are unloaded at the same time |
| 23.09.12 | 2023.09.12 | 1. Support for mounting specified directory of us3 bucket; 2. Fixed the issue where non-root users in the container may not be able to access files created by the us3 console |
| 23.07.24 | 2023.08.01 | Support for RSSd Cloud Disk on bare metal |
| 22.09.1 | 2022.09.17 | 1. Dynamic scheduling of RSSD cloud disks to solve RSSD Cloud Disk Mounting Problem |
| 21.11.2 | 2021.11.22 | 1. csi udisk supports Arc mode; 2. csi udisk supports specifying business group; 3. Optimized us3 mounting parameters |
| 21.11.1 | 2021.11.04 | 1. Adapted to the situation where s3fs returns success but the actual mounting fails； 2. Fixed the mounting failure caused by the change in the length of the US3 public-private key pair； 3. Always operate the mount through the node us3lancher service |
| 21.09.1 | 2021.09.07 | Move s3fs mount operation to Node, Please note that upgrading to this version will cause the pod mount point of US3/UFile to be invalid； |
| 21.08.1 | 2021.08.12 | Optimized the scheduling mechanism of CSI plug-ins to avoid being evicted |
| 21.07.1 | 2021.07.05 | The cloud disk supports bare metal |
| 21.04.1 | 2021.04.28 | Expose UDisk related parameters in Kubelet Metrics |
| 21.03.1 | 2021.03.15 | Solved the problem that volume attachment is not deleted after the node is deleted, leading to storage unloading |
| 21.01.1 | 2021.01.13 | The starting size of UDisk has been changed to 1GB |
| 20.10.1 | 2020.10.14 | Support CSI to limit the maximum number of volumes that can be mounted on a node Avoid the possible concurrent competitive behavior caused by the SurferCloud API client |
