Serverless AWS Alerts Plugin

A Serverless plugin to easily add CloudWatch alarms to functions

Installation

npm i serverless-plugin-aws-alerts

Usage

service: your-service
provider:
  name: aws
  runtime: nodejs4.3

custom:
  alerts:
    stages: # Optionally - select which stages to deploy alarms to
      - production
      - staging

    dashboards: true

    nameTemplate: $[functionName]-$[metricName]-Alarm # Optionally - naming template for alarms, can be overwritten in definitions

    topics:
      ok: ${self:service}-${opt:stage}-alerts-ok
      alarm: ${self:service}-${opt:stage}-alerts-alarm
      insufficientData: ${self:service}-${opt:stage}-alerts-insufficientData
    definitions:  # these defaults are merged with your definitions
      functionErrors:
        period: 300 # override period
      customAlarm:
        description: 'My custom alarm'
        namespace: 'AWS/Lambda'
        nameTemplate: $[functionName]-Duration-IMPORTANT-Alarm # Optionally - naming template for the alarms, overwrites globally defined one
        metric: duration
        threshold: 200
        statistic: Average
        period: 300
        evaluationPeriods: 1
        datapointsToAlarm: 1
        comparisonOperator: GreaterThanOrEqualToThreshold
    alarms:
      - functionThrottles
      - functionErrors
      - functionInvocations
      - functionDuration

plugins:
  - serverless-plugin-aws-alerts

functions:
  foo:
    handler: foo.handler
    alarms: # merged with function alarms
      - customAlarm
      - name: fooAlarm # creates new alarm or overwrites some properties of the alarm (with the same name) from definitions
        namespace: 'AWS/Lambda'
        metric: errors # define custom metrics here
        threshold: 1
        statistic: Minimum
        period: 60
        evaluationPeriods: 1
        datapointsToAlarm: 1
        comparisonOperator: GreaterThanOrEqualToThreshold

Multiple topic definitions

You can define several topics for alarms. For example you want to have topics for critical alarms
reaching your pagerduty, and different topics for noncritical alarms, which just send you emails.

In each alarm definition you have to specify which topics you want to use. In following example
you get an email for each function error, pagerduty gets alarm only if there are more than 20
errors in 60s

custom:
  alerts:

    topics:
      critical:
        ok:
          topic: ${self:service}-${opt:stage}-critical-alerts-ok
          notifications:
          - protocol: https
            endpoint: https://events.pagerduty.com/integration/.../enqueue
        alarm:
          topic: ${self:service}-${opt:stage}-critical-alerts-alarm
          notifications:
          - protocol: https
            endpoint: https://events.pagerduty.com/integration/.../enqueue

      nonCritical:
        alarm:
          topic: ${self:service}-${opt:stage}-nonCritical-alerts-alarm
          notifications:
          - protocol: email
            endpoint: alarms@email.com

    definitions:  # these defaults are merged with your definitions
      criticalFunctionErrors:
        namespace: 'AWS/Lambda'
        metric: Errors
        threshold: 20
        statistic: Sum
        period: 60
        evaluationPeriods: 10
        comparisonOperator: GreaterThanOrEqualToThreshold
        okActions:
          - critical
        alarmActions:
          - critical
      nonCriticalFunctionErrors:
        namespace: 'AWS/Lambda'
        metric: Errors
        threshold: 1
        statistic: Sum
        period: 60
        evaluationPeriods: 10
        comparisonOperator: GreaterThanOrEqualToThreshold
        alarmActions:
          - nonCritical
    alarms:
      - criticalFunctionErrors
      - nonCriticalFunctionErrors

SNS Topics

If topic name is specified, plugin assumes that topic does not exist and will create it. To use existing topics, specify ARNs or use Fn::ImportValue to use a topic exported with CloudFormation.

ARN support

custom:
  alerts:
    topics:
      alarm:
        topic: arn:aws:sns:${self:region}:${self::accountId}:monitoring-${opt:stage}

Import support

custom:
  alerts:
    topics:
      alarm:
        topic:
          Fn::ImportValue: ServiceMonitoring:monitoring-${opt:stage, 'dev'}

SNS Notifications

You can configure subscriptions to your SNS topics within your serverless.yml. For each subscription, you'll need to specify a protocol and an endpoint.

The following example will send email notifications to me@example.com for all messages to the Alarm topic:

custom:
  alerts:
    topics:
      alarm:
        topic: ${self:service}-${opt:stage}-alerts-alarm
        notifications:
          - protocol: email
            endpoint: me@example.com

You can configure notifications to send to webhook URLs, to SMS devices, to other Lambda functions, and more. Check out the AWS docs here for configuration options.

Metric Log Filters

You can monitor a log group for a function for a specific pattern. Do this by adding the pattern key.
You can learn about custom patterns at: http://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/FilterAndPatternSyntax.html

The following would create a custom metric log filter based alarm named barExceptions. Any function that included this alarm would have its logs scanned for the pattern exception Bar and if found would trigger an alarm.

custom:
  alerts:
    definitions:
      barExceptions:
        metric: barExceptions
        threshold: 0
        statistic: Minimum
        period: 60
        evaluationPeriods: 1
        comparisonOperator: GreaterThanThreshold
        pattern: 'exception Bar'
      bunyanErrors:
        metric: bunyanErrors
        threshold: 0
        statistic: Sum
        period: 60
        evaluationPeriods: 1
        datapointsToAlarm: 1
        comparisonOperator: GreaterThanThreshold
        pattern: '{$.level > 40}'

Note: For custom log metrics, namespace property will automatically be set to stack name (e.g. fooservice-dev).

Custom Naming

You can define custom naming template for the alarms. nameTemplate property under alerts configures naming template for all the alarms, while placing nameTemplate under alarm definition configures (overwrites) it for that specific alarm only. Naming template provides interpolation capabilities, where supported placeholders are:

• $[functionName] - function name (e.g. helloWorld)
• $[functionId] - function logical id (e.g. HelloWorldLambdaFunction)
• $[metricName] - metric name (e.g. Duration)
• $[metricId] - metric id (e.g. BunyanErrorsHelloWorldLambdaFunction for the log based alarms, $[metricName] otherwise)

Note: All the alarm names are prefixed with stack name (e.g. fooservice-dev).

Default Definitions

The plugin provides some default definitions that you can simply drop into your application. For example:

alerts:
  alarms:
    - functionErrors
    - functionThrottles
    - functionInvocations
    - functionDuration

If these definitions do not quite suit i.e. the threshold is too high, you can override a setting without
creating a completely new definition.

alerts:
  definitions:  # these defaults are merged with your definitions
    functionErrors:
      period: 300 # override period
      treatMissingData: notBreaching # override treatMissingData

The default definitions are below.

definitions:
  functionInvocations:
    namespace: 'AWS/Lambda'
    metric: Invocations
    threshold: 100
    statistic: Sum
    period: 60
    evaluationPeriods: 1
    datapointsToAlarm: 1
    comparisonOperator: GreaterThanOrEqualToThreshold
    treatMissingData: missing
  functionErrors:
    namespace: 'AWS/Lambda'
    metric: Errors
    threshold: 1
    statistic: Sum
    period: 60
    evaluationPeriods: 1
    datapointsToAlarm: 1
    comparisonOperator: GreaterThanOrEqualToThreshold
    treatMissingData: missing
  functionDuration:
    namespace: 'AWS/Lambda'
    metric: Duration
    threshold: 500
    statistic: Average
    period: 60
    evaluationPeriods: 1
    comparisonOperator: GreaterThanOrEqualToThreshold
    treatMissingData: missing
  functionThrottles:
    namespace: 'AWS/Lambda'
    metric: Throttles
    threshold: 1
    statistic: Sum
    period: 60
    evaluationPeriods: 1
    datapointsToAlarm: 1
    comparisonOperator: GreaterThanOrEqualToThreshold
    treatMissingData: missing

Additional dimensions

The plugin allows users to provide custom dimensions for the alarm. Dimensions are provided in a list of key/value pairs {Name: foo, Value:bar}
The plugin will always apply dimension of {Name: FunctionName, Value: ((FunctionName))}
For example:

    alarms: # merged with function alarms
      - name: fooAlarm
        namespace: 'AWS/Lambda'
        metric: errors # define custom metrics here
        threshold: 1
        statistic: Minimum
        period: 60
        evaluationPeriods: 1
        comparisonOperator: GreaterThanThreshold
        dimensions:
          -  Name: foo
             Value: bar

'Dimensions': [
                {
                    'Name': 'foo',
                    'Value': 'bar'
                },
            ]

Using Percentile Statistic for a Metric

Statistic not only supports SampleCount, Average, Sum, Minimum or Maximum as defined in CloudFormation here, but also percentiles. This is possible by leveraging ExtendedStatistic under the hood. This plugin will automatically choose the correct key for you. See an example below:

definitions:
  functionDuration:
    namespace: 'AWS/Lambda'
    metric: Duration
    threshold: 100
    statistic: 'p95'
    period: 60
    evaluationPeriods: 1
    datapointsToAlarm: 1
    comparisonOperator: GreaterThanThreshold
    treatMissingData: missing

License

MIT © A Cloud Guru