CDN configuration not working with the latest version (CDN 2.17)

vsl

New Member
Hi, we are setting up new Azure VMs with latest Ubuntu and have issue with basic setup of the CDN.

WCS 5.3.466 CDN (1 Origin + 2 Edges). Edges log that they connect to the Origin as a CDN node, but "no stream routes ever form": the Origin never registers the Edges, `show cdn-nodes` is empty on the Origin, there is "no persistent TCP connection on port 8084", and any attempt to play an Origin-published stream from an Edge fails with:
WARN MediaHandler - CDN no route for stream <streamName>
StreamStatusEvent: status=FAILED, info="Session does not exist"

We have ruled out the network/firewall (port 8084 is reachable in both directions, verified).

Environment

- **WCS build:** 5.3.466 on all three nodes (Origin upgraded 5.3.146 → 5.3.466; both Edges are fresh 5.3.466 installs)
- **OS:** Ubuntu 24.04.4 LTS
- **JDK:** OpenJDK 21.0.11
- **Hosting:** Microsoft Azure VMs, all in the **same VNet/subnet `172.28.8.0/24`**
- **NAT note (important):** each node has an Azure **public IP that is NOT bound to the NIC** (the NIC only has the private IP). So `ip` = public, `ip_local`/`cdn_ip` = private.

CDN configuration (`flashphoner.properties`)
Origin:

cdn_enabled=true
cdn_role=origin
cdn_ip=172.28.8.4
cdn_nodes_timeout=60
ip=...
ip_local=172.28.8.4

Edge (e.g. Edge 1):
cdn_enabled=true
cdn_role=edge
cdn_ip=172.28.8.5
cdn_point_of_entry=172.28.8.4
cdn_nodes_timeout=60
ip=...
ip_local=172.28.8.5

Symptoms

1. Edge startup — appears to connect:
INFO CDNTransport - main Listening CDN on address streaming-edge2.internal.cloudapp.net/172.28.8.5 port 8084
INFO CdnRouteMonitor - main Start monitoring remote node: 172.28.8.4
INFO CdnHost - main New node 172.28.8.4(ORIGIN) connected

2. But no persistent CDN connection exists. On BOTH the Edge and the Origin:
ss -tn | grep :8084 → (0 established connections)
The Origin is listening (`ss -ltn` shows `172.28.8.4:8084`), but nothing stays connected.

3. The Origin never registers the Edges.** Grepping all logs (current + rotated) on the Origin for `New node` / `CdnHost` / the edge IPs (172.28.8.5, 172.28.8.6) returns **nothing**. `show cdn-nodes` (CLI) and the CDN REST endpoints return **empty** on the Origin.

## Network verification (rules out firewall)

Port 8084 reachability tested with raw TCP (`/dev/tcp`) between nodes:

| From → To | Result |
|-----------|--------|
| Edge → Origin private (172.28.8.4:8084) | **OPEN** |
| Origin → Edge private (172.28.8.5:8084) | **OPEN** |

Additional logs or ssh access can be provided if needed. Thanks!
 
Good day.
We reproduced the issue in our test environment. You should set
Code:
cdn_nodes_timeout=-1
(the default value) on all the nodes.
We raised the ticket WCS-5026 to investigate the side effect of non-default values of this parameter.
 
Last edited:
Hey, we run into an issue while testing. WebRTC stream is published on origin and can be played on origin, but when trying to play it on the edges we get an error after 30s:
Stream shutdown by auditor's timeout

Setup is the same as written previously: origin + 2 edges within the same private subnet. Playing from both edges results with the same error.

Claude analysis is attached, but it does not add much additional info.
 

Attachments

Please check the following:
1. Publishing keyframe interval should be stable. For example, in OBS settings it should not be 0 (auto):
1785368307774.png

2. WebRTC/UDP ports (30000-33000 by default) should be opened in private network between origin and edges
If this does not help, reproduce the issue and collect a reports as described both on origin and edge where it's reproducing. Then send the reports archives using this private form. We will check the settings and logs both for origin and edge servers.
 
About the ticket WCS-5026. We fixed the cdn_nodes_timeout > -1 issue in build 5.3.469. Please update and check.
About the playback issue. There is the ticket WCS-4953 that may affect the CDN playback case when SRT publisher stops and starts again. As workaround, try the following:
1. Reduce SRT publishing timeouts (timeout and maxTimeStampDiff parameters)
Code:
streams:
  srt1:
    port: 42002
    transport: srt
    timeout: 1000
    hasAudio: true
    hasVideo: true
    maxTimeStampDiff: 1000
This should be set in mpegts_ingest.yml file
2. And reduce the CDN session activity tmeout
Code:
wcs_activity_timer_timeout=5000
This shoud be set in flashphoner.properties file

Please also note that the subscriber must reconnect to the stream if SRT publisher stops publishing then starts again.
 
Good day.
About the ticket WCS-4953. Since build 5.3.470 a viewers may stay connected to the reserved SRT stream: when publisher accidentaly stops streaming, subscribers will see a last frame played (new subscribers do not see anything), and when publisher resumes streaming, the stream should continue to play for all subscribers. The timeout options recommended above may be rolled back to their defaults.
 
Top