Synchronization issues on DU side using kubernetes setup

[lhms1234]

Issue Description

• UE can't connect to the network because the synchronization is periodically lost
• RU packet error counter increasing
• When gNB runs on bare metal, the setup works as expected. But when the gNB runs in kubernetes, the synchronization issue occurs
• Both linuxptp methods works in bare metal: linuxptp on bare metal and on kubernetes

Setup Details

• Server

  • 5GC: Open5GS in kubernetes
  • CU / DU: srsRAN (commit 4cf7513e9fe988449aadedce3278063de761a4ac) container in kubernetes
  • HW configuration:
  • Processor: Intel(R) Xeon(R) Gold 6348 CPU @ 2.60GHz
  • RAM: 256 GB
  • Kernel: 5.15.0-1037-realtime
  • OS: Ubuntu server 22.04.4 LTS

• RU: Foxconn RPQN 7801E (v3.1.12q.551)

• UE: Amarisoft UE (version 2022-03-19)

• Switch Falcon-RX/812/G xHaul & PTP Grandmaster

• RF

  • BW: 50 MHz (133 PRBs)
  • SCS: 30 KHz
  • Band: n77 (TDD)
  • Frequency: centered in 3825 MHz
  • MIMO 4x4

• srsRAN helm version: 7e19cef73b2f0c9b38f324bb61a9379c8a50f94a

Expected Behavior

Amarisoft UE should connect to the network.

Actual Behaviour

Amarisoft UE loses the synchronization.

Steps to reproduce the problem

Considering that the k8s cluster was alteready installed using kubespray and the recommended optimizations was done (realtime kernel enabling, O-FH MTU configured to 9000, etc - except by the DPDK usage), the tutorial was followed.

Additional Information

Sample of the warning messages from O-FH: gnb_ofh_sample.log RU log: ru.log Amarisoft UE log: ue-export.log.zip srsRAN gNB values file: values.yaml.txt Wireshark capture (NGAP, E2AP): host_capture.pcap.txt Pods: pods.log

[pgawlowicz]

@lhms1234 any updates on this issue?