Repository navigation
NethVoice: Review UDP/TCP Port Assignment #7498
Description
As more NethVoice instances are deployed on a single NethServer 8 (NS8) node, it becomes increasingly important to ensure that:
- UDP port usage does not conflict across applications (check all service RTP, Janus, etc.)
- TCP port usage does not conflict across applications (check all service LDAP phonebook, etc.)
-
Review UDP Port Assignments:
- Perform a detailed audit of how UDP ports are allocated for RTP/media and SIP signaling for Asterisk and Janus in each NethVoice instance.
- Determine the maximum number of coexisting NethVoice applications that can be safely hosted on a single NS8 node without port overlap.
-
Review TCP Port Assignments:
- Perform a detailed audit of how TCP ports are allocated for LDAP service for example in each NethVoice instance.
- Determine the maximum number of coexisting NethVoice applications that can be safely hosted on a single NS8 node without port overlap.
Activity
[-]NethVoice: Review UDP Port Assignment and Limit Concurrent Calls[/-][+]NethVoice: Review UDP/TCP Port Assignment and Limit Concurrent Calls[/+]tommaso-ascani commentedon May 8, 2026
Test Cases — Managed RTP/UDP ports (PR nethesis/ns8-nethvoice#817)
This PR requests a 2003-port UDP block + 36 TCP ports from ns8-core and
derives every consumer from it. SIP UDP is distinct from SIP TCP.
+0 … +999+1000 … +1999+2000+2001+2002Update baselines — a temporary fix (1.6.8, commit
4202d5ad) made Satellitereuse
SATELLITE_HTTP_PORTand unset the dedicatedSATELLITE_RTP_PORT; this PRre-introduces it at
+2000. Updates must migrate correctly from both origins:SATELLITE_RTP_PORT(old model)SATELLITE_HTTP_PORTreuse;SATELLITE_RTP_PORTunsetEnvironment: ns8-core cluster, 2 nodes for clone/move; one user domain +
one NethVoice Proxy per node; root SSH.
Images:
<IMG>= PR image; update baselines1.7.4and1.6.7.Placeholders:
<ID>= module id (e.g.nethvoice1) ·<NODE>= node id.Automated tests (Robot Framework —
tests/)Already covered by the automated suites; no manual run needed.
ASTERISK_SIP_UDP_PORT≠ASTERISK_SIP_PORTrc=0, no abortManual tests
Each test is self-contained: do the step, run the command, check the result.
A1 — Clean install
add-module <IMG> <NODE>→ note<ID>. →rc=0.rc=0.runagent -m <ID> systemctl --user --failed→ no failed units.A13 — Ports are correct after install
runagent -m <ID> env | grep -E 'ASTERISK_SIP_PORT|ASTERISK_SIP_UDP_PORT|ASTERISK_IAX_PORT|SATELLITE_RTP_PORT|ASTERISK_RTPSTART'→ all set;
ASTERISK_SIP_UDP_PORT≠ASTERISK_SIP_PORT.runagent -m <ID> podman exec freepbx asterisk -rx 'pjsip show transports'→ the UDP transport is bound to
ASTERISK_SIP_UDP_PORT.ss -lun | grep <ASTERISK_SIP_UDP_PORT>→ the SIP UDP port is listening.B1 — Two instances on the same node don't collide
add-module <IMG> <NODE>→<ID2>; configure it.journalctl _UID=$(id -u <ID2>) | grep -i "address already in use"→ no matches (includes DUNDI 4520 and Janus demo 5002/5102).B2 — Removing an instance frees its ports
remove-module <ID2>sqlite3 /var/lib/nethserver/node/state/ports.sqlite "SELECT * FROM UDP_PORTS;"→ no rows left for<ID2>.Update tests
Before each update, set a marker to check persistence: create a test extension in the PBX and note the admin password.
C1 — Update from 1.7.4 (current stable)
1.7.4instance.runagent -m <ID> env | grep SATELLITE_RTP_PORT→ empty (unset).api-cli run update-module --data '{"module_url":"<IMG>","instances":["<ID>"],"force":true}'rc=0.runagent -m <ID> env | grep SATELLITE_RTP_PORT→ now set (Satellite has its own port again).runagent -m <ID> podman exec freepbx asterisk -rx 'pjsip show transports'→ UDP transport on the newASTERISK_SIP_UDP_PORT, no manual restart needed.rc=0, nothing changes (idempotent).C2 — Update from 1.6.7 (before the temporary fix)
1.6.7instance.runagent -m <ID> env | grep SATELLITE_RTP_PORT→ set (old dedicated port).<IMG>(same command as C1 step 3). →rc=0.runagent -m <ID> env | grep -E 'SATELLITE_RTP_PORT|UDP_PORTS_RANGE'→ Satellite port is inside the new UDP block; no stale old value.C3 — Update next to an existing instance (collision)
1.7.4or1.6.7) with overlapping old ranges.<IMG>.UDP_PORTS_RANGEof the two instances → non-overlapping.journalctl _UID=$(id -u <ID>) | grep -i "address already in use"→ no matches.C4 — Update to a wrong image fails cleanly
api-cli run update-modulewith a non-existentmodule_url. → fails; the instance keeps working on the current image.D1 — Clone to another node
api-cli run clone-module --data '{"module_id":"<ID>","node":2,"replace":false}'→rc=0. Note<ID_DST>on node 2; configure it.runagent -m <ID_DST> podman exec freepbx asterisk -rx 'pjsip show transports'→ UDP transport on the destination
ASTERISK_SIP_UDP_PORT, TCP transport on the destinationASTERISK_SIP_PORT— UDP ≠ TCP.runagent -m <ID_DST> env | grep UDP_PORTS_RANGE→ different range from the source.D2 — Clone with FreePBX offline (config rewrite)
E1 — Move to another node
api-cli run clone-module --data '{"module_id":"<ID>","node":2,"replace":true}'→rc=0.runagent -m <ID_DST> env | grep MODULE_UUIDand the source → no duplicate UUID; source removed.runagent -m <ID_DST> podman exec freepbx asterisk -rx 'pjsip show transports'→ on the destination SIP UDP/TCP ports.F1 — Backup and restore
api-cli run cluster/run-backup --data '{"id":<repo>}'→rc=0.api-cli run restore-module --data '{"repository":"<uuid>","path":"<path>","snapshot":"","node":<NODE>,"replace":false}'→rc=0(does not abort).runagent -m <ID> podman exec freepbx asterisk -rx 'pjsip show transports'→ SIP UDP/TCP on the restored instance's ports.runagent -m <ID> systemctl --user status satellite→active (running)(Satellite PostgreSQL restored).F2 — Restore in place (replace=true)
"replace":true. →rc=0; instance works; ports correct.G1 — FreePBX restart keeps ports
runagent -m <ID> systemctl --user restart freepbxrunagent -m <ID> podman exec freepbx asterisk -rx 'pjsip show transports'→ SIP UDP still onASTERISK_SIP_UDP_PORT.G2 — Reconfigure doesn't drop calls
configure-moduleagain changing a non-port setting (e.g. an integration).runagent -m <ID> podman exec freepbx asterisk -rx 'pjsip show transports'→ ports unchanged; Asterisk not restarted unnecessarily.H1 — Remove frees everything
remove-module <ID>sqlite3 /var/lib/nethserver/node/state/ports.sqlite "SELECT * FROM UDP_PORTS;"andTCP_PORTS→ no rows for<ID>.J1 — Real call (final gate)
Notes
allocator capacity limit depend on ns8-core — verify at integration level.
[-]NethVoice: Review UDP/TCP Port Assignment and Limit Concurrent Calls[/-][+]NethVoice: Review UDP/TCP Port Assignment[/+]22 remaining items
edospadoni commentedon Oct 12, 2026
Testing release:
Amygos commentedon Oct 12, 2026
Release
nethesis/ns8-nethvoice1.7.10