TKG nodes experience spontaneous reboots (kernel panics).
This issue is observed on nodes running CNF workloads.
Symptoms include:
3.2, 3.3, 3.4
Photon OS5 kernel 6.1.75-1.ph5
The issue is a known upstream Linux kernel bug in the Transparent Huge Pages (THP) management logic.
Specifically, during a PMD split (__split_huge_pmd), the kernel executes BUG_ON(!pte_none(*pte)).
The panic occurs because the kernel expects a newly created Page Table Entry (PTE) to be empty (zeroed), but it detects residual data, triggering an invalid operation (ud2).
The fix requires upgrading the Photon OS kernel to a version that includes the upstream backport.
Identify Target Kernel: The fix is integrated into Photon kernel version 6.1.143-1.ph5 or later.
Log in to the VMware Telco Cloud Automation user interface.
Go to Administration > Configurations.
Click Infrastructure Requirements.
To add the kernel version, click ADD KERNEL VERSION and provide the following information:
Name: Select the kernel name.
Version: Enter the kernel version.
Click Add.
Make the necessary changes to the CSAR to add the kernel by logging into the TCA Manager UI.
From the Network Function section under Catalog, select the Catalog Name.
Click EDIT and go to Resources.