Feed health you can verify.

Follow a quote through the system

CoinPriceFeeds keeps separate timing information for the venue and the local processing path. That makes it possible to distinguish:

• a source that is already delayed before the quote arrives;
• a slowdown inside the pricing engine; and
• a delay between a calculated quote and the client connection.

Those measurements are kept by source or client feed where appropriate, rather than collapsed into one number that cannot identify the responsible stage.

Know whether each feed is active

Monitoring tracks the amount of market data received, the number of client-facing quotes produced, and the last time each feed delivered output.

It also records connected consumers by feed. An operations team can answer “Is the client connected and receiving?” without waiting for a screenshot from the other side.

Source cancellations, processing errors, and quote-protection events are counted as separate signals. A feed can be generally busy while one input or one symbol class is failing; the monitoring model is designed to reveal that partial degradation.

Detect unusual quiet periods

Simple absolute thresholds are weak for markets with daily sessions and weekends.

CoinPriceFeeds can compare current activity with the equivalent period in an earlier week. That provides a more useful baseline for a feed that is naturally quieter at some times and very active at others.

Availability, latency, application error, and server resource conditions can also trigger alerts. Operational notifications can include a rendered chart so the person receiving the alert has context on a phone.

Alert thresholds and notification channels are set for the client and deployment during onboarding.

Use one operational dashboard

The monitoring view combines infrastructure and feed behaviour:

• server CPU, memory, disk, network, and file-descriptor state;
• incoming and outgoing quote rates;
• source, processing, and send delay;
• connected clients and last output age;
• source invalidations and application errors; and
• the pricing-rule fingerprint for every feed.

Prometheus metrics and Grafana access can be made available by arrangement when a client wants to integrate feed health into its own operations process.

Compare redundant endpoints directly

The cross-server comparison tool connects to primary and backup feeds at the same time. It checks more than whether both sockets open.

For each symbol it can report:

• price and spread differences;
• activity-rate mismatch;
• liveness on one endpoint but not another;
• invalid or crossed output; and
• configuration-fingerprint agreement.

Thresholds can be chosen for an onboarding test, routine health check, or release gate. Structured output lets the result feed an automated process.

Read more about redundant price-feed design.

Let clients verify from their side

CoinPriceFeeds can provide a small reference streaming client for connection checks, a traffic-logging proxy for parser disputes, and the endpoint-comparison tool.

This keeps verification independent of the service-side monitoring view. A technical team can check what actually arrives at its network edge.

The authenticated client dashboard and quote-state snapshot provide another view of current values, age, status reasons, and source relationships.

Monitor meaning, not only volume

High quote volume does not prove that the feed is correct. Low volume does not automatically mean it is broken.

Useful monitoring combines:

• activity — are quotes moving where expected?
• timeliness — how old are they at each stage?
• validity — are sources being cancelled or outputs rejected?
• consumption — are downstream clients connected and receiving?
• agreement — do primary and backup produce comparable results?
• configuration — are both endpoints using the intended pricing rules?

That is the level at which a brokerage can make a confident operational decision.