This page lists expected behaviors, known limitations, and workarounds for the SC//HyperCore™ HyperVisor and for guest operating systems.
SC//HyperCore Expected Behavior
SC//HyperCore updates may include CPU microcode updates and changes to virtualized CPU features. If a CPU feature is not available to the whole cluster, or is being deprecated, VMs that can use that feature need a power cycle. This keeps the whole cluster reliable. Affected VMs get a tag that shows they need a power cycle. Until the power cycle is done, these VMs may have trouble with live migration.
Starting in SC//HyperCore 9.6, the SC//HyperCore software requires TLS 1.3 for REST API HTTPS requests. TLS 1.2 is deprecated to improve security. This change may affect scripts, older PowerShell versions on Windows Server 2019 and earlier, and third-party integrations.
Updating to SC//HyperCore version 9.1.24 or later deprecates SSL certificates that are less than 2048 bits. If no custom certificate is in place, the update creates a new default certificate. If you use a custom certificate, make sure it is greater than 2048 bits. Otherwise, the system update fails during the pre-check phase. It won’t go ahead until the certificate meets the new standard.
You can’t take a snapshot while a node in the cluster is offline. Snapshots stay blocked until all nodes have rejoined the cluster.
At certain points in an upgrade, shutting down a VM from inside the guest can affect progress. Don’t change VM state from inside the guest while an upgrade is running.
Replication from SC//HyperCore versions before 9.2 is not supported. If you have questions about replicating from systems running SC//HyperCore 9.1 or earlier, contact Scale Computing Support.
Two-port NUC-based nodes have issues with DHCP assigning leases over LAN0. To address this, the default network setup for these nodes assigns LAN1 as primary. Failover to LAN0 brings the DHCP issues back.
Guest Operating Systems
Disconnecting the performance drivers from CentOS 5 blocks incoming connections. To restore inbound traffic, run a service network restart on the VM.
On FreeBSD 10, 11, and 12, the guest OS doesn’t see live-added virtual drives until you reboot.
Very rarely, Windows 10 may crash on startup during boot. Reboot the VM to fix the issue.
Windows drives sometimes show 0 as the serial ID. This value is cosmetic and doesn’t affect the system in any way.
Windows 11 may display an alert that “Memory Integrity is disabled.” Enabling this setting results in severe performance degradation in Windows 11 VMs running on SC//Platform™. Scale Computing recommends keeping this setting disabled.
Windows 11 and Server 2022 may show Error 27 in the Windows Event Log, which points to TPM issues. This shouldn’t affect performance or cause issues with the TPM implementation.
SC//HyperCore Software
Alerts
Squelched alerts are logged in the cluster log. You typically see them during events such as a rolling upgrade. In these cases, CRITICAL or WARNING level conditions are created but not shown to the end user.
The SC//HyperCore software no longer sends email alerts for snapshot quiesce failures. Extensive internal analysis found that these alerts were generated excessively and gave no actionable insight into SC//HyperCore issues.
SC//HyperCore Web Interface
If you connect to the SC//HyperCore web interface through a node that has lost its backplane connection, the login prompt hangs indefinitely or shows a “Disconnected…” message that never resolves. In this case, connect through an online node.
The Network activity gauge on a VM card may show activity when the guest OS shows none, or when the NIC is disconnected. The web interface correctly reports all traffic sent to the VM’s vnet interface. This traffic can reach the interface whatever the NIC connection state, and whether or not the VM itself is sending network traffic. Excessive broadcast traffic shows as network activity, but the vnet interface drops it before it reaches the guest.
Snapshots and Replication
The initial replication of cloned VMs doesn’t show progress in the SC//HyperCore web interface. During this initial replication, the web interface reports “transmission idle” even though replication is active.
After you delete a remote cluster connection for replication, you’re prompted to create a new remote connection, even if another one already exists. To avoid this prompt when you have more than one remote cluster connection, refresh the browser after you delete a connection. Then set up the new replication.
Deleting a remote cluster connection stops the snapshot schedule of any VM that was still replicating to it. Those VMs take no more snapshots, local or replicated, until you assign them a snapshot schedule that doesn’t replicate. Replication itself can’t resume until the remote connection is re-established and replication is set up on the VM again. Pause or cancel replication on all affected VMs before you delete the connection.
The initial replication of a VM clone can’t be cloned. The next snapshot can be cloned.
Replication doesn’t alert you when it fails to create new snapshots. It sends an alert only after a snapshot has not been replicated for 24 hours.
Virtualization
If you replicate or import a VM to a cluster with a smaller resource footprint, the cloned VMs can end up with higher CPU and RAM settings than the cluster supports. Before you start cloned VMs, edit them to bring RAM and CPU within the target cluster’s supported limits.
Windows VMs may log Windows Event ID 122 in the system log, with the message “access to drivers on Windows Update was blocked by policy.” To prevent this event, install the VirtIO serial drivers from the CD Drive Scale Tools installation wizard.
Very rarely, when you remove a drive, the hypervisor fails to delete it within the allotted time. To fix this, fully shut down the affected VM and then restart it.
Windows 2012 Standard (non-R2) can crash when you live-add new drives one after another, or when you live-delete a drive. To prevent this, wait 30 seconds before you add a new drive. If the crash happens, reboot the VM to fix it.
VMs that use nested virtualization may crash or hang during live migration. Power down nested virtualization VMs before system updates.
UEFI VMs freeze or crash if you add or remove a disk while the VM is running. Power down UEFI VMs before you add or remove disks. Power cycle affected VMs to restore functionality.
Scale Guest Agent Tools may have trouble installing the right drivers and running the Guest Agent on Windows VMs. The guest agent may fail to start, and some drivers may not install properly. For help with Guest Agent issues on SC//HyperCore systems, contact Support.