Bug ID 2162705: Tmm restarting on multi-NUMA AWS instances with ENA interfaces

Last Modified: Feb 03, 2026

Affected Product(s):
BIG-IP All, Install/Upgrade(all modules)

Known Affected Versions:
17.5.1.3

Fixed In:
21.0.0.1, 17.5.1.4

Opened: Nov 26, 2025

Severity: 1-Blocking

Symptoms

Tmm is in the restart loop because dpdk driver is failing to attach with the error message in tmm log: notice dpdk: [0000:00:06.0]: Multiple NUMA nodes usage is unsupported.

Impact

Unable to use dpdk driver on some large AWS instances.

Conditions

- BIG-IP VE large instance deployed on AWS cloud. - NUMA node count more than 1 (check "lscpu | grep NUMA").

Workaround

Switch to sock driver: https://my.f5.com/manage/s/article/K10142141

Fix Information

DPDK correctly initializes the memory on multi-NUMA AWS instances.

Behavior Change

Guides & references

K10134038: F5 Bug Tracker Filter Names and Tips