threshold-metrics

Creates new metrics from NR timeslice metrics for outlier / threshold analysis. Provide a timeslice metric name, and a threshold. Dimensional metrics will be generated that show how many data points are above the threshold.

To build

go build

To run

Set the following environment variables:

export NEW_RELIC_ACCOUNT=YOUR_ACCOUNT_ID
export METRIC_NAME=METRIC_TO_QUERY
export METRIC_WHERE=NRQL_WHERE_CLAUSE
export METRIC_FACET=FACETS_TO_INCLUDE
export THRESHOLD=TARGET_THRESHOLD
export NEW_RELIC_LICENSE_KEY=YOUR_LICENSE_KEY
export NEW_RELIC_USER_KEY=YOUR_USER_API_KEY

You can optionally set the polling interval (default 1m):

export POLL_INTERVAL=5m

Let's look at a demo example NRQL query:

SELECT apm.mobile.ui.thread.duration, action, scope FROM Metric
WHERE appName = 'Acme Telco -Android' AND action LIKE '%onCreate'

The above query would be specified as follows:

export METRIC_NAME=apm.mobile.ui.thread.duration
export METRIC_WHERE="appName = 'Acme Telco -Android' AND action LIKE '%onCreate'"
export METRIC_FACET=action,scope

The METRIC_NAME specifies which field contains the timeslice aggregate values. The METRIC_WHERE is applied as a NRQL WHERE clause to filter to specific entities and metrics. The METRIC_FACET specifies which fields (entity.guid is automatically included) to apply as attributes on the dimensional metrics generated.

Then set a THRESHOLD x1, which will be used to find the area shown below: For example,

export THRESHOLD=0.5

Then you can run as follows:

./threshold-metrics

This will generate a dimensional metric, calculating the count over threshold in a normal distribution, to each 1-minute timeslice metric aggregate.