KB4960 Veeam Appliance on AMD EPYC: boot panic

Call: Hold for Veeam Software Appliance 13.1+ virtual machines on AMD EPYC hosts until you have a recovery plan. On Nov 4, Veeam published KB4960: after a reboot (including one that follows a kernel update), the appliance can hit an early-boot kernel panic and stay down until you power the VM off and on again. A guest restart is not enough.

What breaks: Boot fails with:

Kernel panic - not syncing: jitterentropy: Initialization failed with host not compliant with requirements: 9

The Linux kernel in the appliance runs a jitterentropy health test in FIPS mode. On some AMD EPYC CPUs (Veeam names 7262, 7302, and 7F72) that test fails intermittently and the kernel stops. Veeam says a kernel fix is not available for the version the appliance ships, and the issue is in the kernel, not in Backup & Replication itself once the guest is up.

Who’s hit: Veeam Software Appliance 13.1 and later as a VM on those AMD EPYC hosts. Once the appliance has booted successfully, normal Backup & Replication operation is unaffected until the next reboot that trips the test.

What to do:

• Prefer redeploying the appliance onto a host with a different processor if you can.
• If you stay on EPYC: disable automatic updates, install updates manually, and watch the VM through the reboot.
• If the panic hits: power off the VM at the hypervisor, then power it on. Do not rely on a soft reboot of the guest.

Sources: Veeam KB4960 (opens in a new tab)↗⁠.

Sources

• https://www.veeam.com/kb4960⁠↗ (opens in a new tab)