Awesome

tsuru-autoscale

Features

• HTTP api to manage auto scale configuration
• integration with tsuru
• Wizard that makes easy configure an auto scale for tsuru applications
• Support HTTP based data sources
• Support HTTP based actions
• Support alarm scripts in javascript

Concepts

The tsuru-autoscale is based on three elements: data sources, actions and alarms.

Data sources

Data source is a http endpoint that provide the data to an alarm. Is based on the data source data that the alarm will execute an action.

Actions

Action is a http endpoint that is called when the alarm expression result is true.

Alarms

Alarm is composed by data sources, actions and by an expression. When the expression result is true the actions will be executed.

Wizard

Wizard is an easy way to use autoscale with tsuru. Wizard creates the alarms for scale up and scale down, based on simple inputs like: ``

Install as tsuru application

Create tsuru app using Go platform

tsuru app-create autoscale go

Configuring MongoDB

We should use environment variables to configure the database:

tsuru env-set "MONGODB_URL=mongodb://172.17.0.1:27017/tsuru_autoscale" -a autoscale

Deploy the applications

tsuru env-set GO_PKG_PATH=github.com/tsuru/tsuru-autoscale -a autoscale

tsuru app-deploy . -a autoscale

API Reference

list data sources

curl <autoscale-url>/datasource

add data source

curl -XPOST -d '{}' -H "Content-Type: application/json" <autoscale-url>/datasource

remove a data source

curl -XDELETE <autoscale-url>/datasource/{name}

list actions

curl <autoscale-url>/action

add an action

curl -XPOST -d '{}' -H "Content-Type: applicaiton/json" <autoscale-url>/action

remove an action

curl -XDELETE <autoscale-url>/action/{name}

list alarms

curl <autoscale-url>/alarm

add an alarm

curl -XPOST -d '{}' -H "Content-Type: application/json" <autoscale-url>/alarm

remove an alarm

curl -XDELETE <autoscale-url>/alarm/{name}

Configuring Wizard to works with tsuru

To wizard works fine with tsuru it is necessary to configure some data sources and the actions to scale up and scale down.

Add the scale up action

curl -XPOST -d '{"name": "scale_up", "url": "http://<tsuru_url>/apps/{app}/units", "method": "PUT", "body": "units={step}&process={process}", "headers": {"Authorization": "bearer <tsuru-token>", "Content-Type": "application/x-www-form-urlencoded"}}' -H "Content-Type: application/json" <autoscale-url>/action

Add the scale down action

curl -XPOST -d '{"name": "scale_down", "url": "http://<tsuru_url>/apps/{app}/units?units={step}&process={process}", "method": "DELETE", "headers": {"Authorization": "bearer <tsuru-token>", "Content-Type": "application/x-www-form-urlencoded"}}' -H "Content-Type: application/json" <autoscale-url>/action

Add data source to get the number of units

curl -XPOST -d '{"name": "units", "url": "http://<tsuru_url>/apps/{app}", "method": "GET", "headers" : {"Authorization": "bearer <tsuru_token>"}}' -H "Content-Type: application/json" <autoscale-url>/datasource

Add data source to get cpu data from ElasticSearch

Only configure it if you are using ElasticSearch as tsuru metrics backend.

curl -XPOST -d '{"name": "cpu", "url": "http://<elasticsearch_url>/<elasticsearch_index>/cpu_max/_search", "method": "POST", "body" : "{\"size\":0, \"query\": {\"filtered\": {\"filter\": {\"bool\": {\"must\": [{\"range\": {\"value\": {\"lt\": 500}}},{ \"term\": {\"app.raw\": \"{app}\"}}, {\"term\": {\"process.raw\": \"{process}\"}}]}}}}, \"aggs\": {\"range\": {\"date_range\": {\"field\": \"@timestamp\", \"ranges\": [{\"from\": \"now-5m/m\", \"to\": \"now\"}]}, \"aggs\": {\"date\": {\"date_histogram\": {\"field\": \"@timestamp\", \"interval\": \"1m\"}, \"aggs\": {\"max\": {\"max\": {\"field\": \"value\"}}, \"avg\": {\"avg\": {\"field\": \"value\"}}}}}}}}", "public": true}' -H "Content-Type: application/json" <autoscale-url>/datasource

Add data source to get cpu data from Prometheus

Only configure it if you are using Prometheus as tsuru metrics backend.

curl -XPOST -d '{"name": "cpu", "url": "<prometheus_url>/api/v1/query?query=max(irate(container_cpu_system_seconds_total{container_label_tsuru_process_name=\"{process}\",container_label_tsuru_app_name=\"{app}\"}[1m]))*100", "public": true}' -H "Content-Type: application/json" <autoscale-url>/datasource