xOTTWorkers, placement and delivery
Separate processing location from public delivery, inspect node health and add remote VOD capacity.
On this page
Main panel and workers
The main panel owns accounts, configuration, orchestration, the database and subscriber access. Workers execute assigned media jobs and report telemetry. Each worker needs its own applicable worker license and a compatible release. A worker is not another independent main panel.
| Control | What it decides |
|---|---|
| Placement | Where the stream process runs. |
| Delivery mode | Whether viewer media is relayed by the controller or served from a signed worker origin. |
| VOD destination | Where a VOD job downloads/processes/stores its file. |
| Proxy / DNS | Which supported upstream requests use regional routing. |
Add and verify a worker
- Open Operations → Servers & Workers and add the target host, SSH details and worker license. Verify the target before submitting installation.
- Follow the queued/installation progress and log. An installer finishing is not proof of a current heartbeat.
- Wait for online status and run the node health check.
- Name the worker clearly, set groups and conservative capacity limits, then place one test stream.
- Confirm processing location, public playback and telemetry.
Placement precedence
Per-stream overrides take precedence over service defaults and the global policy. Choose inherited/automatic placement, main-only, a specific worker, eligible workers or a group according to the UI. A strict pin with “fail if unavailable” behaves differently from main-server fallback. Restart/recovery should preserve your explicit saved placement.
Resource limits affect new placement; they do not create network shaping. Drain mode excludes a worker from new starts. Do not remove a busy worker without reviewing active streams, VOD jobs and any evacuation/recovery operation.
Relay versus direct origin
The public entry URL can stay on the main panel in both modes. In Controller relay, media bandwidth passes through the main server. In Direct origin, an entry request can issue a signed redirect and the assigned origin serves subsequent media. Inspect actual HTTP responses rather than concluding from the first URL that all bytes pass through the controller.
Read worker observations
Use the inline worker overview for concise specs, stream/session counts and resource usage; use Monitoring for detailed diagnosis. Offline, disabled, draining and stale are different states. A stale worker must not be counted as healthy merely because its last known CPU was low. Network interfaces and limits must reflect the actual media NICs.
Worker VOD storage
A native worker VOD destination moves download and processing as well as storage. Configure its folder from VOD → Storage or the worker’s folder control. Mount disks before choosing paths. See the worker-storage recipe. Live stream placement does not automatically choose the destination for every VOD download.
Common failure paths
- Heartbeat absent: check worker service, controller reachability, token and versions.
- Heartbeat present but commands fail: check main-to-worker network and authenticated API reachability.
- Direct playback fails: verify advertised origin, TLS, firewall and chosen Direct Failure policy.
- Unexpected main placement: inspect effective stream/service/global policy, eligibility and explicit fallback.
- After update: ensure the worker actually reports the same release, not just that an update was queued.