English for New Relic Observability
Learn the English vocabulary for discussing New Relic APM data, entities, and alert conditions when investigating performance issues with a team.
New Relic packages a lot of concepts under one dashboard — application performance monitoring, infrastructure metrics, and alerting — and conversations about “what New Relic is showing” often skip past which specific concept is actually being discussed. Learning to name entities, transactions, and alert conditions precisely helps a team move from “something looks off” to a concrete diagnosis faster.
Key Vocabulary
Entity — any monitored thing in New Relic’s model — an application, a host, a database, a Kubernetes cluster — each with its own health status, golden metrics, and relationships to other entities. “The checkout service entity is showing elevated error rate, but its upstream database entity looks completely healthy, so the problem is probably in the application layer.”
Transaction — a single request or unit of work traced through an instrumented application, broken down into the time spent in each internal function, external call, or database query. “I traced the slow transaction and found it’s spending eighty percent of its time waiting on an external payment API call, not in our own code.”
Golden signal — one of the core metrics New Relic highlights for an entity’s health, typically covering throughput, error rate, latency, and saturation. “Before diving into traces, I always check the golden signals first — in this case, error rate spiked exactly when latency did, which usually points to timeouts.”
NRQL (New Relic Query Language) — a SQL-like query language for querying New Relic’s stored telemetry data directly, used to build custom charts and alert conditions. “Instead of relying on the default dashboard, I wrote an NRQL query to break down error rate specifically by customer tier.”
Alert condition — a rule defined against a metric or NRQL query that triggers a notification when a threshold is breached over a configured time window, forming the basis of New Relic’s alerting policies. “We tightened the alert condition on this service’s error rate because the old threshold was so loose that a real problem could run for twenty minutes before anyone got paged.”
Common Phrases
- “Which entity is actually unhealthy here — the app, the database, or something upstream?”
- “Can you trace one of the slow transactions and see where the time is actually going?”
- “Let’s check the golden signals before we dig into individual traces.”
- “I’ll write a quick NRQL query to break this down by region.”
- “Is this alert condition too sensitive, or did we miss a real spike?”
Example Sentences
Diagnosing a performance regression: “The transaction traces show most of the added latency is in a single external API call, not in our own database queries, so the fix is probably on their side, not ours.”
Explaining an alert: “This alert condition fires when error rate on the checkout entity exceeds two percent for five straight minutes, which is why it didn’t page us for the brief one-minute blip earlier.”
Building a custom view: “I wrote an NRQL query that groups slow transactions by customer segment, since the default golden-signal dashboard doesn’t break performance down that way.”
Professional Tips
- Name the specific entity you’re looking at, not just “New Relic,” when reporting a problem — “the checkout service entity” is actionable, “New Relic is showing errors” is not.
- Trace an individual transaction before assuming a broad slowdown — it often isolates the problem to one external call or query instead of the whole request path.
- Check the golden signals as a first pass before writing custom queries — they’re designed to surface the four things most likely to explain a health change quickly.
- When proposing an alerting change, describe it in terms of the alert condition’s threshold and time window, not just “make it more sensitive” — precise language avoids a policy that’s either too noisy or too slow to fire.
Practice Exercise
- Explain, in one sentence, what an entity represents in New Relic’s model.
- Describe how tracing an individual transaction can reveal whether a slowdown is caused by your own code or an external dependency.
- Write two sentences explaining to a teammate why an alert condition didn’t fire during a brief latency spike.