HealthThresholds
Opt-in thresholds that make the engine emit DEGRADED / RECOVERED status
changes from the per-connection health it already measures — for clients that
want an event, not numbers. Every field is optional; an unset threshold never
trips. Behind/ahead are driven by the smoothed delivery rate (only once
settled), NOT the relative realtimeDelta. See docs/srt-input-metrics.md.
Field |
Type |
Repeated |
Description |
max_delivery_rate |
ahead_of_realtime when settled delivery_rate exceeds this (e.g. 1.1). |
||
min_delivery_rate |
behind_realtime when settled delivery_rate falls below this (e.g. 0.9). |
||
stalled_ms |
stalled when last_media_age_ms exceeds this (progressing:false also stalls). |
||
loss_ratio |
lossy (SRT) when received_lost/received over a sample exceeds this. |
||
dropped_per_second |
lossy (SRT) when the unrecovered dropped-packet rate exceeds this. |
||
latency_over_rtt_min |
latency_below_rtt (SRT) when latency_over_rtt falls below this (aim >= 3-4). |
||
min_duration_ms |
hysteresis: a condition must hold this long before DEGRADED fires (default 0). |
||
clear_after_ms |
hysteresis: a condition must be clear this long before RECOVERED fires (default 0). |
Member of
Message |
Description |
Configure an RTMP Input Media Node. Should be sent once, after the node has been created. |
|
Takes over one accepted connection of an SRT listener running
CONNECTION_OWNERSHIP_PER_CONNECTION. Create it after accepting the
connection (SrtConnectionAccept on the listener’s stream); the two are
unordered on the wire, so a claim arriving first waits for its accept. The
node then owns that connection: the data path, its outbound context, TS
context/tables, timecode, stream statistics, GOP structure, SRT stats and
connection health all arrive on THIS stream, and closing the node closes
the connection. The node lives exactly as long as the connection: on
disconnect it emits |
|