Transaction tracing extends the New Relic service to the tiniest transaction detail levels. Primarily, New Relic only collects streaming metric data. This data is aggregated, so that if the ORM Layer (such as ActiveRecord or Hibernate) was invoked 100 times during a 60 second time period, New Relic would collect one metric that was an aggregate of the 100 invocations.
With the Transaction Tracing feature, just the opposite is true. Transaction tracers collect data for your slowest individual HTTP requests each minute, all the way down to the database queries. If a transaction takes longer than a configurable time period, New Relic captures the exact invocation pattern and details related to the invocation, such as the exact query statement that was issued.
Access to this feature depends on your subscription level.
Recording specific transactions
New Relic records only one transaction trace per harvest cycle (one minute). The slowest transaction during this period is the only one kept.
If you do not see a transaction trace for a particular transaction you know is taking longer than your transaction trace threshold, it may not be the slowest one. If you want to make sure New Relic records a specific transaction, set it up as a key transaction.
To set up the Transaction Tracing feature: From the New Relic APM menu bar, select Applications > (selected app) > Settings > Application. For more information, see Configuring transaction traces.
To view transaction trace records, do either of the following:
From the New Relic APM menu bar, select Applications > (selected app) > Monitoring > Overview, and then select the Web transactions view. Then, from the Transactions section, select a link for any app server transaction trace.
- From the New Relic APM menu bar, select Applications > (selected app) > Monitoring > Transactions. Then select a link for any app server transaction trace.
For more information, see Viewing transaction traces.
The transaction trace's Summary shows each segment of the trace as colored bars. Red does not necessarily mean there is a problem; it simply indicates that a significant amount of time was spent in that segment. You will need to determine a "normal" or "typical" value for the process in order to evaluate whether a red area on the bar chart actually indicates a problem.
For more help
Additional documentation resources include: