flask-cloudwatch-metric

A Flask extension to provide easy and efficient integration with CloudWatch to publish application metrics.

Sample usage

app = Flask(__name__)
metrics_reporter = CloudWatchMetricsReporter(app)

@app.route('/')
def main():
    response = # ... Your application logic
    metrics_reporter.add_metric(
        metric_name="RequestCount",
        value=1,
        unit="Count",
        dimensions=[]
    )
    return response, 200

How it works?

flask-cloudwatch-metric is built on top of boto3 client and Flask framework, to provide efficient per-request metrics publication to AWS CloudWatch.

When a CloudWatchMetricsReporter object is created with a Flask app, the constructor will inject an interceptor(or called filter if you are familiar with JAXRS) into the application. The interceptor runs for every request providing the following functionalities:

• before each request: initialize a in-memory buffer to store metrics for this request
• after each request: publish all the metrics in buffer to CloudWatch
• during the request: metrics are added to the buffer by calling the report API's(e.g. add_metric)

To support running Flask in multi-threading mode, we need to store the in-memory buffer in a place that is separated for each request. To achieve this, we can use the application context provided by Flask's g context(https://flask.palletsprojects.com/en/2.0.x/api/#flask.g). Using the g context, each request can store its own metric objects without interfering with others.