Skip to content

Latest commit

 

History

History
150 lines (110 loc) · 7.32 KB

operation.md

File metadata and controls

150 lines (110 loc) · 7.32 KB

Operation

Day-to-day use of the network. Almost everything is done from the dashboard (http://<core machine IP>:30998).


1. Administrator login

Viewing the state is open. Actions (start, stop, switch core, edit the configuration) need the administrator password:

  • 🔒 Admin, at the top: one login covers every tab and window of the same browser, for 8 h;
  • 🔓 Admin · Sign out ends the session;
  • the password is in sudo cat /opt/open5gs-k3s/state/dashboard-admin-password (core machine).

The EN/PT button at the top switches the dashboard language (remembered by the browser).

2. After powering the machines on

For safety, only the dashboard comes up by itself (CORE_AUTOSTART=false). The core and the gNB are started by the operator.

# Where What to do Ready when
1 dashboard 🔒 Admin → Control tab → check the Active core → ▶ Start all every step ✓ (AMF listening on N2, SMF associated with the UPF)
2 RAN machine wait for the GNSS: journalctl -u ts2phc -f (or the GPS line of the Status button) lines with offset … s2 (up to ~30 min after the boot)
3 RAN machine if the clock jumped (normal after a boot): sudo systemctl restart ptp4l-gm phc2sys phc2sys with an offset of a few ns
4 O-RU reboot the O-RU (it synchronized with the old clock) RU tab: SYNCHRONIZED
5 dashboard gNB (Kubernetes) card: choose the .yml of the active core → ▶ Start the tab checks pod running, DU connected and N2 on the AMF
6 dashboard Overview phones registered with an IP

If the RAN machine reboots, the gNB pod comes back without starting the gNB and shows "stopped (node rebooted)". This is the boot protection: start it again from the dashboard after steps 2–4. See components/ocudu.md.

The local-network IPs (DHCP) can change after a reboot. Only the address used to open the dashboard changes; the cluster and the fiber use fixed IPs.

3. Start and stop

Control tab:

Action What happens
▶ Start all brings the active core up in order (MongoDB → NRF → … → UPF → SMF → AMF → UERANSIM), checking each step in the logs
■ Stop all scales the core and UERANSIM to 0 (nothing is deleted: configuration and subscribers stay)
per-function buttons start / stop / restart one function; restarting the UPF also restarts the SMF, and the AMF restarts the UERANSIM gNB
gNB card start with a .yml / restart / stop the OCUDU gNB

Before powering the machines off, stop the gNB and then the core.

4. Switch core

Available with free5GC installed. Disconnects everyone connected during the switch.

  1. Control tab → Active core card → click the other core.
  2. Leave restart the OCUDU gNB with ticked and choose the .yml that points to the AMF of the new core (the dashboard only lists those).
  3. Switch now. The dashboard stops the gNB → stops the current core → marks the new one → brings the new one up, checking each step → restarts the gNB.

Typical times: Open5GS → free5GC ~1 min; free5GC → Open5GS ~5 min. Every tab then follows the active core (functions, logs, subscribers, sessions, traffic).

Always use the Active core card. Stopping one core and starting the other with the separate buttons does not touch the gNB, which keeps pointing to the old AMF.

Tutorial: Switch core.

5. The OCUDU gNB

From the dashboard (gNB (Kubernetes) card) or from the command line, on the core machine:

./ran/ocudu/gnb.sh configs                 # available .yml files and the AMF of each one
./ran/ocudu/gnb.sh start <yml>             # start (without extension; default: the current one)
./ran/ocudu/gnb.sh stop                    # SIGINT to the gNB, like Ctrl+C
./ran/ocudu/gnb.sh status                  # replicas, .yml, pod, node, restarts
./ran/ocudu/gnb.sh log -f                  # gNB output (full log: /tmp/gnb.log on the node)
Changed… Do
a .yml in ~/ocudu/configs/ (RAN node) restart the gNB. The pod reads the host folder; no rebuild needed
the OCUDU source commit → rebuild the image → ./ran/ocudu/install.sh → restart. See tutorial 7
the core AMF restarted restart the gNB (it does not redo N2 by itself)

6. Subscribers (SIMs)

Core Where
Open5GS dashboard → Core → Subscribers (WebUI :30999); or SUBSCRIBERS in config.env + sudo ./install.sh --deploy-only; or ./scripts/subscribers.sh
free5GC dashboard → Core → Subscribers (WebConsole :31000); or ./free5gc/install.sh --import-open5gs

The keys (K/OPc) never go into git and never show on the screen during the import. Tutorial: Phone on the network.

7. Change the configuration

Where Apply with Effect
dashboard → Settings tab (edits config.env with a diff and validation) the command the dashboard shows saving does not apply
config.env — core/UERANSIM sudo ./install.sh --deploy-only recreates the functions that changed (disconnects users)
config.env — dashboard sudo ./install.sh --dashboard-only only the dashboard restarts
config.env — host/k3s sudo ./install.sh may restart k3s
free5gc/free5gc.env ./free5gc/install.sh recreates free5GC
ran/ocudu/ocudu.env ./ran/ocudu/install.sh updates the Deployment (keeps the .yml and the on/off state)

Be careful:

  • run --deploy-only with Open5GS as the active core: Helm sets the Open5GS replicas back to 1;
  • do not run an installer and a core switch at the same time;
  • after recreating the AMF, restart the gNB.

Tutorial: Change PLMN, slices and IPs.

8. Test

Test Command (core machine) Expected
Open5GS + UERANSIM ./test.sh (-v shows logs on failure) 7 checks OK
free5GC + UERANSIM ./free5gc/test.sh 6 ok, 0 failure(s)
real gNB dashboard → RU and Radio SYNCHRONIZED, DU Ready, ~90 % on time; UEs with throughput
phone dashboard → Overview registered, session with an IP from the active core pool

The tests only read the state and use the simulated UE; the tested core must be running.

9. Useful commands

kubectl get nodes -o wide                         # both machines Ready?
kubectl get pods -A -o wide                       # everything, with the node of each pod
kubectl -n open5gs logs deploy/open5gs-amf -f     # AMF live
kubectl -n ocudu logs deploy/ocudu-gnb -f         # gNB
journalctl -u open5gs-k3s-gnb-ctl -f              # gNB control agent (core machine)
less logs/install.log                             # installer history

Problems: troubleshooting.md.