Doppler is a Clojure service for creating composite Cloudwatch metrics and alarms.
Cloudwatch is a great tool for performing ad-hoc analysis and creating alarms for dev teams. Unfortunately, its alarms are hindered by the inability to specify multiple input metrics and apply conditional logic. At RJMetrics, we have run into a few scenarios where composite Cloudwatch metrics come in handy. For example:
- When an SQS queue is backed up, but the consumers are not de-queueing messages;
- Verifying that the number of messages through various services match up exactly;
- Or, creating aggregate traffic metrics across horizontally scaled clusters.
Build an uberjar with lein uberjar. Right now you need to provide AWS credentials in the local environment, like:
$ AWS_ACCESS_KEY_ID=<access key id> AWS_SECRET_ACCESS_KEY=<secret access key> java -jar doppler.jarYou'll need to customize the config file in order to generate your own metrics. Here's an example config:
;; list of metric inputs
{:metrics [{:namespace "AWS/SQS"
:name "NumberOfMessagesReceived"
:period 300
:dimensions [{:name "QueueName" :value "my-queue"}]
:start-time (time/minus (time/now) (time/hours 1))
:end-time (time/now)}
;; :namespace is the Amazon namespace for the input metric, usually
;; begins with "AWS/".
{:namespace "AWS/SQS"
;; :name is the name of the metric in Cloudwatch.
:name "NumberOfMessagesSent"
;; :period is the period, in seconds, for which you'd like to pull
;; this metric. must be a multiple of 60.
:period 300
;; :dimensions apply filters to the metrics that you pull.
:dimensions [{:name "QueueName" :value "my-queue"}]
;; :start-time is a joda-time object representing the query start time.
:start-time (time/minus (time/now) (time/hours 1))
;; :end-time is a joda-time object representing the query end time.
:end-time (time/now)}]
;; :mapper is a Clojure function that is applied to the initially pulled Cloudwatch
;; data. The inputs to this function look like:
;;
;; {:time <time>
;; :average <average value>
;; :sum <sum value>}
;;
;; This mapper will take the sum value from the most recent data point.
:mapper (fn [i] (-> i last :sum))
;; :reducer is a Clojure function that is applied to the outputs of :mapper. This
;; could be a multi-arity Clojure function, or a 2 input function called by `reduce`.
:reducer -
;; :output defines metadata for the output metric. These values are equivalent to the
;; values provided in the :metrics collection above.
:output {:namespace "SQS"
:name "ReceivedMinusSent"
:unit "Count"
:dimensions [{:name "QueueName" :value "my-queue"}]}