PDU session drops and RACH procedures from non-existing UEs
Dear OAI developers,
I was testing Bartosz' RSSI-threshold branch and it seems to fix our power control issues related to railing the O-RU's ADC. However, we have noticed some remaining issues that might originate from higher layers that I wanted to report here. I was experiencing the following issues:
- PDU sessions are sometimes dropped / released, especially after throughput-heavy traffic (e.g. benchmarks with iperf3) when the UE goes back to idle
- UE loses data connection
- Shortly putting the UE into flight mode and reconnecting the UE again typically solves the issue
- Once the softmodem even crashed after such a PDU drop
- RACH procedures from non-existing UEs
- L2 frequently reports unsuccessful RACH procedures, but there is no other active UE in vicinity of the 5G system, unless the one that is already connected.
- Sometimes, L2 logs even show BLER+SINR statistics from non-existing RNTIs with very low SINR reports
I have attached below some L1 and L2 log files that show the reported issues (filenames indicate corresponding issues). I was using NVIDIA Aerial cuPHY 24-3 for L1, OAI branch "rssi-threshold-with-logs" for L2, and "oai-cn5g-fed" for the CN. The hardware platform is NVIDIA GH200, the UE is a Samsung Galaxy S23 with the latest firmware, and the O-RU is a Foxconn RPQN 7800E.
Please let me know if you need any additional information.
Thank you in advance for your assistance.
Kind regards,
Reinhard Wiesmayr
oai_pdu_session_release-L2_crash_after_throughput_heavy_traffic-lots_of_rach.log
phy_pdu_session_release-L2_crash_after_throughput_heavy_traffic-lots_of_rach.log
phy_speedtest_non_existing_RNTIs_in_L2_log.log
oai_speedtest_non_existing_RNTIs_in_L2_log.log