Multple srsUEs random access failure

[improperaffo]

Dear all,
we are trying to stress test the gNB using multiple srsUEs. When we reach around 15 srsUEs connected, some of them don't fully connect, this is what we can see in their logs:

Active RF plugins: libsrsran_rf_zmq.so
Inactive RF plugins: 
Reading configuration file /opt/srsran/share/ue.conf...

Built in Release mode using commit ec29b0c1f on branch master.

Opening 1 channels in RF device=zmq with args=tx_port=tcp://*:2701,rx_port=tcp://10.43.242.207:2700,base_srate=11.52e6
Supported RF device list: zmq file
CHx base_srate=11.52e6
Current sample rate is 1.92 MHz with a base rate of 11.52 MHz (x6 decimation)
CH0 rx_port=tcp://10.43.242.207:2700
CH0 tx_port=tcp://*:2701
Current sample rate is 11.52 MHz with a base rate of 11.52 MHz (x1 decimation)
Current sample rate is 11.52 MHz with a base rate of 11.52 MHz (x1 decimation)
Waiting PHY to initialize ... done!
Attaching UE...
Closing stdin thread.
Random Access Transmission: prach_occasion=0, preamble_index=0, ra-rnti=0x39, tti=174
Random Access Transmission: prach_occasion=0, preamble_index=37, ra-rnti=0x39, tti=334
Random Access Transmission: prach_occasion=0, preamble_index=55, ra-rnti=0x39, tti=494
Random Access Transmission: prach_occasion=0, preamble_index=62, ra-rnti=0x39, tti=654
Random Access Transmission: prach_occasion=0, preamble_index=48, ra-rnti=0x39, tti=814
Random Access Transmission: prach_occasion=0, preamble_index=37, ra-rnti=0x39, tti=974
Random Access Transmission: prach_occasion=0, preamble_index=48, ra-rnti=0x39, tti=1134

We have tried throttling down in the GNU Radio script, but that did not help. The UEs that do not connect are always different, and sometimes they all connect successfully. We have tried setting a delay between the start up of each UE, to avoid congestions.

What could be the reason for these issues?

[pgawlowicz]

The issue will be fixed in the next release. You can apply this patch and test again.
ra_error_reset_harq.patch.txt

[improperaffo]

Thank you Piotr