- Ya
- 27 Agosti 2026, 16:20 UTC
- Mwandishi
- Kamo
- Ahadi ya
- 6258e92
An endpoint can now REGISTER from the public internet and Kazoo's API reports it: **************** 200 OK via 47.181.8.87:5060. Four separate faults, each of which presented as something other than itself: RabbitMQ was missing **************** Kazoo's registrar_shared_listener declares its exchange as x-consistent-hash, so the broker closed the channel with 503 COMMAND_INVALID and the listener never bound. Nothing consumed authn.*, Kamailio's auth query came back "312 NO_ROUTE", and a REGISTER simply hung after "100 checking your credentials" — kazoo_apps still listed registrar as running the whole time. The plugin ships in the image, it was only switched off. ecallmgr and FreeSWITCH could not see each other. mod_kazoo registers as freeswitch@<hostname>, which under hostNetwork is the bare node name "k3m1", and Erlang long names (-name) REQUIRE a dotted FQDN — "Hostname k3m1 is illegal". Both sides now use short names (-sname, shortname=true), which accept a dotless host. The symptom was a bare `pang` with nothing logged. The edge is built on 2600Hz's own Kamailio (see that repo). It needed KAZOO_DIST_ALL_INTERFACES, since vm.args pins distribution to loopback and ecallmgr has to reach a node named for the host. The edge readiness probe called pgrep, which the Rocky-based image does not ship, so the pod never went Ready while Kamailio was answering SIP perfectly well. It now uses kamcmd core.uptime, which proves the process ANSWERS rather than merely exists. Removed: the Postgres StatefulSet added earlier as a KZ_DB_MODULE backend. 2600Hz's packages carry db_kazoo.so, so the vendored SQLite default works and the extra database was never wired to anything.