Ivs Diagnostics

IVS Diagnostics

When to use

Any Amazon IVS investigation — channel creation, streaming, playback, chat, recording, low-latency optimization, or real-time stages.

Investigation workflow

Step 1 — Collect and triage

aws ivs list-channels --max-results 20
aws ivs get-channel --arn <channel-arn>
aws ivs list-streams --filter-by health=STREAM_LIVE
aws ivs get-stream --channel-arn <channel-arn>
aws cloudwatch get-metric-statistics --namespace AWS/IVS --metric-name ConcurrentViews --dimensions Name=Channel,Value=<channel-arn> --start-time <start> --end-time <end> --period 300 --statistics Maximum

Step 2 — Domain deep dive

aws ivs list-stream-keys --channel-arn <channel-arn>
aws ivs get-stream-session --channel-arn <channel-arn> --stream-id <stream-id>
aws ivs list-recording-configurations
aws ivs-chat list-rooms --max-results 20
aws ivs-realtime list-stages --max-results 20

Step 3 — Detailed investigation

aws ivs list-stream-sessions --channel-arn <channel-arn> --max-results 10
aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventSource,AttributeValue=ivs.amazonaws.com --max-results 10
aws s3 ls s3://<recording-bucket>/<path>/

Read references/guardrails.md before concluding on any IVS issue.

Gotchas: IVS

• IVS has two modes: Low-Latency Streaming (channels) and Real-Time (stages). Different APIs.
• Stream keys are sensitive — treat like passwords. Regenerate if compromised.
• IVS auto-records to S3 if recording configuration is attached to channel.
• Low-latency mode uses WebRTC for sub-second playback via IVS Player SDK.
• Standard latency uses HLS and works with any HLS player.
• Chat is a separate service (ivs-chat) with its own rooms and tokens.
• IVS is available in limited regions — check availability before creating resources.

Anti-hallucination rules

1. IVS Low-Latency and Real-Time are different products with different APIs.
2. Stream keys must be kept secret — never log or expose them.
3. IVS Player SDK is required for low-latency playback — standard HLS players add latency.
4. Recording requires S3 bucket in the same region as the IVS channel.
5. Chat tokens are short-lived — generate new tokens for each session.
6. Spend no more than 2 minutes on any single hypothesis. Pivot if inconclusive.

14 runbooks