An AWS Cloudwatch Library for Clojure
[org.promotably.clojars/apollo "0.2.0"]
and
[apollo.core :as apollo]
Cloudwatch limits the number of writes you can make over short periods of time. It's best to keep an in-memory collection of metric data that is vacuumed and sent to Cloudwatch on a periodic basis. This library seeks to do that.
Aggregated metric data is accumulated in-memory and then vacuumed and sent to Cloudwatch in batches.
The client creating functions below take these optional keys as arguments:
:credentials AWSCredentials
:provider AWSCredentialsProvider
:config ClientConfiguration
(def my-sync-client (apollo/create-cw-client))
(def my-async-client (apollo/create-async-cw-client))
;; or
(def my-sync-client (apollo/create-cw-client :credentials AWSCredentials :config ClientConfiguration))
(def my-sync-client (apollo/create-cw-client :provider AWSCredentialsProvider ...))Metric recording takes place in the context of a namespace and dimensions. You can call the apollo.core/record! function directly, supplying your own namespace string and dimension map as the first two of the arguments, or get a recording function returned by apollo.core/get-context-recorder which takes a namespace string and dimension map and returns a function wrapped in the namespace and dimensions which can then be called with the metric-name, a value, unit, and optional additional namespace and/or dimension extensions.
(ns my.things.controller
(:require [apollo.core :as apollo]))
(def ns-metric-recorder (apollo/get-context-recorder "my.things.controller" {:dim-name :value})
;; later...
(ns-metric-recorder "requests" 1 :Count)
;; or
(ns-metric-recorder "requests" 1 :Count :dimensions {:endpoint :endpoint-name})
(ns-metric-recorder "errors" 1 :Count :namespace "errors") ;; passing the optional ns extension will reset dimensions to an empty map.
;; calling record! directly
(apollo/record! "my.namespace.string" {:endpoint "endpoint-name"} "requests" 1 :Count)
There are also the following helper functions for simple increment and decrement recording operations:
(apollo.core/get-context-inc-recorder namespace dimensions)
(apollo.core/get-context-dec-recorder namespace dimensions)
The functions returned by these functions only need to be called with the metric name to increment or decrement.
The dimensions map in the above examples can be an empty map.
Simply call apollo.core/enable-sys-metrics! with a namespace prefix string as it's only argument.
A scheduled executor will vacuum the in-memory collection of aggregated metrics and send them to Cloudwatch on the provided schedule.
Get a ScheduledExecutor:
(apollo.core/create-vacuum-scheduler)
Start the scheduler:
(apollo.core/start-vacuum-scheduler! delay-secs interval-secs scheduler client)
"client" is your Cloudwatch client...
Stop the scheduler:
(apollo.core/stop-vacuum-scheduler! scheduler)
For convenience, the apollo.component ns provides an implementation of Stuart Sierra's Lifecycle component. Use it directly or as a reference when creating your own. The component will create the Cloudwatch client, & schedule and manage the vacumming of metrics. To work it requires a map of the following form to be passed in when creating a new instance:
{:apollo {:sys {:enable? true ;; enables system metric collection
:ns-prefix "my-app"}
;; all of the client options below are optional
:client {:credentials AWSCredentials
:provider AWSCredentialsProvider
:config ClientConfiguration
;; or :async
:type :sync}
;; scheduler config is NOT optional
:scheduler {:delay 30
:interval 30}}}The created client and scheduler instances are accessed via the client and scheduler properties of the started component.
Make this documentation better and add some docstrings!
Copyright © 2015 Promotably
Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent