Live primary and backup feeds, with proof they agree.

Primary and backup run at the same time

A useful backup is already processing the same sources and applying the same pricing policy before the primary has a problem.

CoinPriceFeeds can run a client feed simultaneously across multiple servers. The appropriate placement depends on the client’s own infrastructure and network paths; regional endpoints can reduce cross-continent delay and shared points of failure.

Clients may keep both links connected and choose how their platform or bridge handles failover.

Keep pricing configuration in agreement

Every server exposes a compact fingerprint of the rules content it has loaded. The value is available on the stream, in the client dashboard, and through monitoring.

If primary and backup fingerprints differ, the issue is visible before a failover. If they agree, the client has a direct check that both servers are applying the same version of the pricing policy.

The fingerprint is designed to represent the rule content rather than the time it was loaded, so two servers can agree even when their reload happened at different moments.

Compare prices, not only configuration

Matching rules are necessary, but the live result also depends on source state and network timing.

CoinPriceFeeds comparison tooling can connect to every endpoint at once and measure:

• mid-price and spread agreement by symbol;
• tick-rate differences;
• symbols live on one server but quiet on another;
• crossed or otherwise invalid output;
• the final observed values; and
• the loaded rules fingerprint.

The report can be produced in a human-readable form or structured output suitable for an acceptance check.

Use independent connection paths

Redundant servers are most valuable when the client reaches them through paths that do not share every dependency.

CoinPriceFeeds can provide separate hostnames and network endpoints for primary and backup delivery. Domain and DNS paths can also be separated where the client’s continuity plan requires it.

The endpoint arrangement is agreed around the client’s network and continuity requirements.

Preserve warm pricing state

Some pricing functions carry meaningful state: the previous accepted source quote, a persistent fallback value, or a temporary dampening cushion.

CoinPriceFeeds saves relevant pricing state and restores it across routine restarts. This reduces the risk of a visible price jump caused solely by maintenance.

Recover after a process or server restart

Production services start automatically after a server reboot and retry after an unexpected process exit.

Connectors also supervise their own receive loops and reconnect when a session fails or becomes unresponsive. FIX market-data subscriptions are restored after logon.

Automatic recovery does not replace monitoring. The monitoring and verification layer confirms whether sources, feeds, and consumers are actually moving again.

Change configuration without replacing the whole service

Routine configuration work—such as adding a feed, changing a mapping, or updating a venue session—can be fetched and validated while the existing configuration continues serving.

Applying an accepted replacement requires a brief feed reconnect. If the replacement cannot start correctly, CoinPriceFeeds restores the previous working setup.

Verify changes before and after release

The standard operating approach compares live endpoints before a deployment and again after it.

That distinction matters. A pre-existing symbol difference should not be mistaken for a new regression, and a new difference should not be dismissed because the feed is generally “up.”

Backups can be updated and checked before primaries so a healthy partner remains available during the change.

Dedicated deployments when needed

Some clients need stronger infrastructure isolation, dedicated capacity, or a separate change boundary. A dedicated deployment can serve only that client’s feeds.

This is a provisioning choice rather than a different pricing product. The rules, dashboard, delivery, and verification model remain consistent.