Distributed Open5GS: Registration Reject [95] and AUSF HTTP 400 during Slicing Handshake
17 views
Skip to first unread message
Yasir Nawaz
May 2, 2026, 3:54:59 AMMay 2
to Powder Users
I am a graduate researcher currently developing a 5G network testbed focused on Network Slicing. My environment consists of an Open5GS core distributed across multiple nodes with UERANSIM providing the gNB and UE. While I have successfully provisioned subscribers in MongoDB and configured specific slices (SST: 1, SD: 0x000001), I am encountering persistent signaling errors during the NAS registration phase that prevent the UE from reaching an MM-REGISTERED state.
Questions:
Q1. Registration Reject [95] & Semantic Mismatches: I am consistently receiving a Registration reject [95] (Semantically Incorrect Message) during the NAS handshake. Given that my open5gs-ue.yaml and MongoDB entry are aligned regarding keys and slice IDs, what hidden semantic mismatches—such as protocol versions or specific IE (Information Element) formatting—typically cause this rejection in a UERANSIM-to-Open5GS setup?
Q2. AUSF HTTP 400 Error during Auth Retrieval: My AMF logs report an HTTP response error [400] immediately after the Setup NF Instance [type:AUSF] step. Since SBI connectivity is confirmed and the NRF shows all NFs as registered, what specific conditions cause the AUSF or UDM to return a 400 "Bad Request" when it attempts to generate authentication vectors for a known SUCI?
Q3. "Unknown UE by SUCI" Despite Provisioning: The AMF log reports Unknown UE by SUCI for IMSI 999700000000001. I have verified the MCC/MNC (999/70) and Routing Indicator match across all configs. Are there known issues with how the UDR parses the subscriber database that could lead to a lookup failure, specifically regarding MNC digit length (2 vs 3 digits) or string-to-integer conversion in MongoDB?
Q4. Slice Selection Failure & Context Release: When requesting a specific slice using a hex-formatted SD (0x000001), the AMF triggers a context release. Beyond basic snssai definitions in the UE and subscriber database, are there additional mandatory configurations required in the nssf.yaml or amf.yaml to ensure the core correctly routes requests to the intended slice in a multi-node environment?
Q5. Hardware Compatibility & Database Stability: During my initial setup, I faced significant hardware compatibility issues while configuring MongoDB on a single node. Now that I am using a distributed architecture, could low-resource hardware or network latency between the UDR and a remote MongoDB instance be contributing to these authentication timeouts or "Unknown UE" errors? Are there recommended hardware minimums for maintaining stable database connectivity in a 5G slicing environment?
Questions:
Q1. Registration Reject [95] & Semantic Mismatches: I am consistently receiving a Registration reject [95] (Semantically Incorrect Message) during the NAS handshake. Given that my open5gs-ue.yaml and MongoDB entry are aligned regarding keys and slice IDs, what hidden semantic mismatches—such as protocol versions or specific IE (Information Element) formatting—typically cause this rejection in a UERANSIM-to-Open5GS setup?
Q2. AUSF HTTP 400 Error during Auth Retrieval: My AMF logs report an HTTP response error [400] immediately after the Setup NF Instance [type:AUSF] step. Since SBI connectivity is confirmed and the NRF shows all NFs as registered, what specific conditions cause the AUSF or UDM to return a 400 "Bad Request" when it attempts to generate authentication vectors for a known SUCI?
Q3. "Unknown UE by SUCI" Despite Provisioning: The AMF log reports Unknown UE by SUCI for IMSI 999700000000001. I have verified the MCC/MNC (999/70) and Routing Indicator match across all configs. Are there known issues with how the UDR parses the subscriber database that could lead to a lookup failure, specifically regarding MNC digit length (2 vs 3 digits) or string-to-integer conversion in MongoDB?
Q4. Slice Selection Failure & Context Release: When requesting a specific slice using a hex-formatted SD (0x000001), the AMF triggers a context release. Beyond basic snssai definitions in the UE and subscriber database, are there additional mandatory configurations required in the nssf.yaml or amf.yaml to ensure the core correctly routes requests to the intended slice in a multi-node environment?
Q5. Hardware Compatibility & Database Stability: During my initial setup, I faced significant hardware compatibility issues while configuring MongoDB on a single node. Now that I am using a distributed architecture, could low-resource hardware or network latency between the UDR and a remote MongoDB instance be contributing to these authentication timeouts or "Unknown UE" errors? Are there recommended hardware minimums for maintaining stable database connectivity in a 5G slicing environment?
Reply all
Reply to author
Forward
0 new messages