1 What is New?

[New] SNMP Service

• SNMP protocol and port can be enabled/disabled at port management.

[New] Online SP service

• Support online SP patching.
• Support 4 types of connectivity to SP server for different scenarios.

[New] Disk-Based Datastore Creation

• Support to create virtual storage volume based on disks instead of nodes.

[New] Storage interface multiplexing 

• Storage interface is now supported to share the same physical interface with others usage of the interface.
• VM will be prioritized to be power on in the healthy node.
• Currently, only Mellanox CX4 and Intel X710 support for four interfaces multiplexing.

[New] PaaS integration

• PaaS is now supported in 6.2.0.
• KubeManager is available to integrate for PaaS management and operation.

[New] Virtual Storage Forecast

• Prediction for the storage capacity usage.
• Prediction for the disk's bad sectors.
• Prediction for the SSD's lifetime.

[Enhanced] VM Scheduling

• Added VM Group and Host group.
• Support VM Group Anti-Affinity.
• Support Host Affinity for selected VM Group.

[Enhanced] Security Aspect

• Password policy allows specifying password length, complexity, and validity.
• Support VM disk encryption.
• A Port management option is available to disable the unused port.
• VM data erasure after the VM had been deleted.

[Removed] Windows Shared Folder

• The backup repository no longer supports adding Windows Shared Folder.
• The existing Windows Shared Folder is still usable but unable to add new.

[Removed] Spare disk

• The spare disk has been removed since HCI6.2.0 in virtual storage.
• The existing spare disk will not be affected while a new spare disk for expanding or initializing virtual storage is no longer available.

2 Resolved Issues

[Fixed] VM data loss when running disksafe application (Severity: High)

• For the windows VM migrated to HCI, data loss might occur when the disksafe application is running for backup.

[Fixed] Memory Reclaiming (Severity: Medium)

• It is fixed that the memory reclaiming features is not enabled by default.

[Fixed] Importing OVF file (Severity: Medium)

• VM import might fail if the OVF file being imported when the VS having high latency.

[Fixed] Failed to take a snapshot for consistency group (Severity: Low)

• When taking a snapshot for the consistency group, the operation failed and prompted System unavailable.

3 Upgrade Instruction

3.1 Confirmation Before Upgrade

Support upgrade from the following earlier HCI/aCloud versions:

• 6.2.0_EN Build20201116
• 6.1.0_EN
• 6.0.1_R1_EN
• 6.0.1_EN
• 6.0.0_R5EN
• 6.0.0_R4EN
• 6.0.0_R3EN
• 6.0.0_EN
• 5.8.8R1_EN
• 5.8.8_EN
• 5.8.7R1_EN
• 5.8.6_EN
• 5.8.5_EN
• 5.8.3_EN

Notes:

1. Support upgrade from versions installed with any service pack but not support upgrade from the custom version.
2. It should be upgraded to version 5.8.3_EN first, to upgrade the version earlier than 5.8.3_EN to this version.
3. It is required to install the package for a pre-upgrade check a week before the official upgrade.

3.2 Upgrade Limitations

1. Before upgrade to this version, the version earlier than 5.8.3_EN should be upgraded to 5.8.3_EN first.
2. As a cluster scales out, converting configuration files takes longer, so does the upgrade process. For reference: upgrading an 11-node cluster running 1000 virtual machines may take 30 minutes, while upgrading a 2-node cluster running 1000+ virtual machines (virtual storage: 6.4 TB) may take 14 minutes.
3. After upgrade, it is required to reboot the device with a graphics card inserted but does not have IOMMU enabled.

Immediate Upgrade of Configurations, Logs, and Data:

• Yes.

Reboot Required After Upgraded

Reboot is required after upgrading from version 5.8.5_EN, all R versions, and earlier versions, while it is not required after upgrading from the following versions:

• 5.8.6_EN
• 5.8.7R1_EN
• 5.8.8_EN
• 5.8.8R1_EN
• 6.0.0_EN
• 6.0.0R3_EN
• 6.0.0R4_EN
• 6.0.0R5_EN
• 6.0.1_EN
• 6.0.1R1_EN
• 6.1.0_EN
• 6.2.0_EN Build20201116

Time Taken:

• 40 minutes.

Upgrade Recommendations:

• Install the package for pre-upgrade check before the upgrade. Make sure there are no hardware issues, and then perform the upgrade.
• It is recommended to perform the upgrade during a non-peak hour or non-production hour.

3.3 Upgrade Recommendations

• Install the package for pre-upgrade check before the upgrade. Make sure there are no hardware issues, and then perform the upgrade.
• Upgrade aCloud/HCI to the latest version first before upgrade aCMP/SCP. Then, make sure aCloud/HCI and aCMP/SCP version are compatible. Starting from SCP6.1.0, it is supported to manage multi-version of HCI.

3.4 Upgrade Procedure

Step 1. Check whether the current version is a custom version or applied with any sp patch. Contact Sangfor Support for further assistance if it is a custom version or applied with any sp patch.

Step 2. Check the recycle bin and empty the recycle bin.

Step 3. Check the Overlay IP Pool. It is required to be configured to avoid any problem caused by the upgrade.

Step 4. Install the package for the pre-upgrade check.

Step 5. Solve the detected issues or contact Sangfor support for further assistance.

Step 6. If the current version is earlier than 5.8.6_EN, it is required to upgrade the version to at least 5.8.6_EN and above first. Before upgrading, virtual network devices (NFV) and the virtual machine is required to be shut down. Besides, the nodes are required to reboot after upgrade successfully.

Step 7. Make sure the version of the update package to be installed is the HCI6.2.0_EN official version.

Step 8. Check the environment for upgrades.

Step 9. Load the update package of the HCI6.2.0_EN version.

Step 10. For hot upgrade, exit upgrade after upgrade finishes. For cold upgrade, restarting the device is required to finish the upgrade.

3.5 Handling of Upgrade Failure

Scenario 1: There is a virtual machine or virtual network device which has not been shut down.

Solution: Shut down virtual machines and virtual network devices manually.

Scenario 2: Network is not stable due to reboot or unstable cluster.

Solution: Wait for the cluster or network to become stable and then perform an upgrade again.

4 Precautions

1. Recommend perform database perform testing with hosts with RAM greater than 256GB.
2. Memory over-commitment:

• It's not recommended unless necessary because memory over-commitment could have a negative impact on performance.
• When the actual consumed memory of VMs is nearing the host's physical memory, a memory swap will happen, and the performance of VMs will be degraded. When SWAP is used up, low priority VMs will be killed off.

3. When migrating/restoring VMware VMs to aCloud/HCI, IP addresses of the VMs can be automatically obtained as long as the source VMs are installed with VMware Tools.
4. A live upgrade is supported by 5.8.6 and later versions.