Failure in creating networking entities on Edge node due to memory constraints

Products

• VMware NSX

Issue/Introduction

• Configuring a networking entity (namespace, VRF etc.) fails on the edge due to memory constraints.
• This can impact running services (routing, LB, DNS etc.) where the service state can end up being 'down'.
• Following log lines in var/log/kern.log on the Edge is an indication for this scenario:
  
  2025-02-10T03:00:36.560Z at033687-vdnet-nsxedge-ob-24541354-461-1250-rtqa22-1 kernel - - - [530198.583669] runc:[1:CHILD]: page allocation failure: order:5, mode:0x40cc0(GFP_KERNEL|__GFP_COMP), nodemask=(nil),cpuset=1bf1dca##########################################abca17f22,mems_allowed=0

Environment

VMware NSX

Cause

This issue happens in corner cases intermittently when kernel is unable to allocate a contiguous block as requested by the networking entity creation system call. The 'available' memory in the system is still high, but the allocation can't proceed as there is fragmentation. The creation eventually succeeds once the defragmentation happens.

Resolution

There is no resolution as the issue is triggered if Edge node runs out of a contiguous block of memory.

Workaround

Enter and exit maintenance mode - this would ensure that the networking entity config creation is re-triggered.