Skip to content

Latest commit

 

History

History
54 lines (52 loc) · 4.01 KB

AWS-lambda-metrics.md

File metadata and controls

54 lines (52 loc) · 4.01 KB
Raw

Metrics for Lambdas

  • AWS Embedded Metrics are an ideal solution for generating metrics for Lambda functions that will track historical data.
  • They are a method for capturing Cloudwatch metrics as part of a logging request.
  • This is good because it avoids the financial and performance cost of making a putMetricData() request.
  • It also makes it easier to find the point at which the metric is updated in both the logs and in the code itself.
  • This does not work at all for our EC2 apps as their logs do not pass through Cloudwatch.
  • This pull request gives a working example of how to embed metrics in your logging request
  • This document gives a good summary of why AWS embedded metrics are so useful
  • Full details can be found in the AWS Documentation, but here are the highlights:
  • To use AWS Embedded metrics, logs must be in JSON format.
  • A metric is embedded in a JSON logging request by adding a root node named “_aws” to the start of the log request.
  • The metric details are defined within this "_aws" node.
  • The following code snippet shows a logging request updating a single metric:
  {"_aws": {  
            "Timestamp": 1574109732004,  
            "CloudWatchMetrics": [  
              {  
                "Namespace": "lambda-function-metrics",  
                "Dimensions": [["functionVersion"]],  
                "Metrics": [  
                  {  
                    "Name": "time",  
                    "Unit": "Milliseconds"  
                  }  
                ]  
              }  
            ]  
  },
  "functionVersion": "$LATEST",
  "time": 100,
  "requestId": "989ffbf8-9ace-4817-a57c-e4dd734019ee"
  }
  • Within the "_aws" node there is a "timestamp" field and a "CloudwatchMetrics" field
  • The "timestamp" field is in epoch milliseconds format
  • The "CloudwatchMetrics" field contains an array of CloudwatchMetric objects
  • Each CloudwatchMetrics object contains the "Namespace", "Dimensions" and "Metrics" fields
  • "Namespace" is a string matching the namespace all metrics in the request are listed under.
  • "Dimensions" is an array of DimensionSet objects.
  • If you do not have any Dimensions you must still define the field and pass an empty nested array "[[]]".
  • The metrics field contains an array of MetricDefinition objects. These contain the "Name" and "Unit" for the metric.
  • The "Name" field of the metric must correspond with the name of a field in the body of the logging request.
    This field will hold the value for the metric.
  • Metrics sent by this request must correspond with the logging objects contained in the request.
  • The value of the metric is not specified in the metric definition,
    instead the metric name is used as a reference to look up the value of the corresponding field in the logging request.
  • Further reading: