Advanced Usage

The telemetry SDK includes components to simplify data aggregation for long running applications.

Batches

Batches provide a standard interface for aggregating and flushing data across different data types. This interface is used by the harvester to forward data from batches to a client.

Batches will automatically track the interval over which data is aggregated so you don’t have to manually set interval_ms on metrics!

All public batch methods are thread safe.

Batches are broken out by data type that they contain:

Data Type	Batch Type
`Metric`	`MetricBatch`
`Event`	`EventBatch`
`Log`	`LogBatch`
`Span`	`SpanBatch`

Example

from newrelic_telemetry_sdk import CountMetric, MetricBatch

metric_batch = MetricBatch()

# Record that there have been 5 errors
metric_batch.record_count("errors", 5)

# Calling flush will clear the batch and reset the interval start time
items, common = metric_batch.flush()

# The interval is automatically set by the batch!
print(common["interval.ms"])

Harvester

A Harvester flushes a batch and sends data through a client at a fixed harvest interval.

The Harvester class is a threading.Thread and has start and stop methods.

Example

The example code assumes you’ve set the following environment variables:

NEW_RELIC_LICENSE_KEY

import atexit
import os
from newrelic_telemetry_sdk import GaugeMetric, MetricBatch, MetricClient, Harvester

metric_client = MetricClient(os.environ['NEW_RELIC_LICENSE_KEY'])
metric_batch = MetricBatch()
metric_harvester = Harvester(metric_client, metric_batch)

# Send any buffered data when the process exits
atexit.register(metric_harvester.stop)

# Start the harvester background thread
metric_harvester.start()

# The data will buffer and send every 5 seconds or at process exit
metric_batch.record_gauge("temperature", 78.6, {"units": "Farenheit"})