Skip to content

Monitoring Agora

Agora monitoring has two parts:

  • a system that collects raw metrics;
  • a system that aggregates and can raise alerts.

Raw data is exposed as metrics over HTTP(S) as JSON or OpenMetrics (Prometheus). Most values are cumulative counters (the usual Prometheus style): bytes, error counts, and similar.

The aggregating system (Agora’s own or the customer’s) is expected to differentiate those counters over time.

We provide alert templates to bootstrap monitoring and extend them for your environment.

Built-in admin monitoring

Besides external metric collection, Agora includes monitoring screens in the admin UI so operators and admins can assess platform state without a separate observability stack.

Streamer monitoring

The streamers page shows aggregated metrics per registered streamer.

The list includes:

  • streamer hostname;
  • online stream count vs. total configured streams;
  • connected clients;
  • CPU load;
  • disk usage.

You can:

  • open a streamer card;
  • refresh data manually;
  • use the page as a quick infrastructure overview.

It helps spot node degradation, overload, or general platform load.

Stream monitoring

Stream monitoring is available on the stream list and inside each stream card.

On the overview list you see:

  • overall stream status;
  • input bitrate;
  • current output bitrate;
  • input operating mode;
  • outgoing publication state.

The list supports manual refresh and auto-refresh on an interval.

Detailed stream statistics

The stream card shows detailed cluster statistics.

Depending on configuration, the UI may show:

  • overall cluster status;
  • primary and backup state in Twincast mode;
  • effective input source;
  • input bitrate;

  • input statistics:

  • bytes;

  • frames;
  • retries;
  • input switches;
  • media info changes;
  • errors;
  • last DTS time;
  • outgoing publication state and stats.

In Twincast, stats are split for primary and backup to simplify troubleshooting per ingest path.

Outgoing publication monitoring

Agora also shows egress / outgoing publication statistics.

Per push you may see:

  • publication URL;
  • push status;
  • bytes sent;
  • frame count;
  • error count.

In Twincast, egress stats may be split across servers so operators can compare primary and backup.

External monitoring

Built-in views are convenient for day-to-day work; for production, connect an external monitoring system.

That enables:

  • long metric history;
  • charts and dashboards;
  • alerting;
  • integration with corporate observability;
  • feeding Prometheus, Zabbix, or other stacks.

What to watch first

For initial production monitoring, commonly track:

  • streamer and relay availability;
  • CPU, disk, and bandwidth load;
  • active streams and clients;
  • input bitrate and input errors;
  • Twincast primary / backup health;
  • outgoing publication errors;
  • recording, archive, and content delivery health.