New Relic APM telemetry gap after collector connectivity loss

1. Does the agent resume harvest cycles automatically after connectivity is restored?

Most current New Relic APM agents are built to retry sending data once the collector endpoint becomes reachable again. The retry logic is triggered by the network layer: when a TCP connection to the collector on port 443 can be established, the agent resumes its harvest cycle without requiring a service restart. The time until the first successful harvest depends on the agent’s internal back‑off algorithm, typically ranging from a few seconds to a minute. However, the exact delay and the maximum number of retry attempts can vary by language and agent version.

2. What retention period does New Relic apply to stale entity metadata before garbage collection?

Entity names and tags are stored as part of the entity’s metadata. New Relic retains this metadata for as long as there is associated telemetry that falls within the account’s data‑retention window. For most accounts, metrics and events are kept for 30 days, traces for 90 days, and logs for 90 days (unless the account has extended retention). Once all telemetry for an entity has aged beyond its retention window, the platform removes the entity from the UI. Therefore, stale entity metadata typically disappears after the longest retention period that applies to any data type the entity has produced.

3. Does renaming a service preserve existing history under the same GUID or create a new entity?

The entity GUID is derived from the combination of the agent’s app_name (or equivalent) and the environment. Changing the application name in the agent configuration causes the agent to report telemetry under a new GUID. The previous entity remains in the UI and continues to hold its historical data until the retention period elapses. Consequently, a rename does not merge history; it creates a new entity. If the same app_name is later restored, the agent will resume reporting to the original GUID.

Likely Explanation for Observed Gaps

When connectivity is lost, the agent’s in‑memory buffer stores unsent telemetry for a limited time. If the outage lasts longer than the buffer capacity, older samples are dropped. Upon reconnection, only the data that survived the buffer is sent, which explains the apparent gap. Because the agent automatically retries, no manual intervention is required to resume reporting.

Diagnostic Detail Needed

To give a precise estimate of the retry interval and buffer size, please provide the language agent and its version that is running in your environment. These parameters influence the exact recovery behavior and the amount of data that can be preserved during a connectivity loss.