dragon.telemetry.telemetry

Dragon’s API to generate and visualize time series metrics in grafana utilizing the telemetry infrastructure

Classes

Telemetry

This class is used to generate and visualize time series metrics in Grafana utilizing the Dragon telemetry infrastructure.

class Telemetry

Bases: object

This class is used to generate and visualize time series metrics in Grafana utilizing the Dragon telemetry infrastructure. The telemetry infrastructure is started when the runtime is started with --telemetry-level set to greater than zero. A --telemetry-level=1 is reserved for user metrics and will not collect any default metrics. The user can add data to the local node database using the add_data method. The data can then be visualized in Grafana or retrieved and analyzed utilizing the AnalysisClient.

Example usage:

import dragon
import multiprocessing as mp
import time
from dragon.telemetry import Telemetry

if __name__ == "__main__":
    dt = Telemetry()

    mp.set_start_method("dragon")
    pool = mp.Pool(10)

    for _ in range(10)
        start = time.time()
        pool.map(f, list(range(100)))
        dt.add_data("map_time", time.time()-start, telemetry_level=2)

    pool.close()
    pool.join()

    dt.finalize()

__init__(metrics_url='http://localhost:4243/api/metrics')

property level

add_data(ts_metric_name: str , ts_data: float , timestamp: int = None, telemetry_level: int = 1, tagk: str = None, tagv: int | str = None) → None 

Adds user defined metric data to node local database that can then be retrieved via Grafana

Parameters:

ts_metric_name (str ) – Metric name used to store data and retrieve it in Grafana. This should be consistent across nodes. Grafana’s retrieval will add the hostname to the metric.
ts_data (float ) – Time-series data point
timestamp (int , optional) – time stamp for time-series data point, defaults to int(time.time())
telemetry_level (int , optional) – telemetry data level for metric. if the data_level is greater than the launch specified telemetry level the data will not be added to the data base, defaults to 1
tagk (str , optional) – tag key for the datapoint
tagv (int | str , optional) – tag value

finalize() → None : Finalize shuts down the telemetry service if it was started. It can be called when the user is done with telemetry. If it is not called the user will have to Ctrl-C from the terminal to shutdown the telemetry service.