# Install On Heroku

This article describes how to install Fluentd (td-agent) on [Heroku](http://www.heroku.com/).

## With non-Treasure Data backend

Heroku doesn't allow users to install separate processes within a single dyno. You will thus need to setup Fluentd as a separate Heroku application. This will become you central log aggregation server.

Treasure Data provides a boilerplate repository to get you started. Follow the steps below to create Fluentd (td-agent) as a Heroku application.

```
# Clone
$ git clone git://github.com/treasure-data/heroku-td-agent.git
$ cd heroku-td-agent
$ rm -fR .git
$ git init
$ git add .
$ git commit -m 'initial commit'

# Create the app & deploy
$ heroku create --stack cedar
$ git push heroku master

# Update your config
$ vi td-agent.conf
$ git commit -a -m 'update config file'

# Deploy
$ git push heroku master
```

## With Treasure Data as a Backend

Since [Treasure Data](http://www.treasuredata.com), the primary sponsor of Fluentd, is a [Heroku Addons provider](https://addons.heroku.com/treasure-data), some of you came here to stream data to Treasure Data through Fluentd.

If

```
          http                          
Your App ------> Heroku-hosted Fluentd ------> Treasure Data
```

is what you want, there are two ways to deploy it.

### 1. Heroku Button

Clicking on the following button automatically launched td-agent on Heroku:

[![](https://www.herokucdn.com/deploy/button.png)](https://heroku.com/deploy?template=https://github.com/treasure-data/heroku-td-agent)

You can read more about this approach [here](https://github.com/treasure-data/heroku-td-agent).

### 2. Using Git

This is similar to "With non-Treasure Data Backend", except you do not need to edit the config. Instead, simply update the configuration variable:

```
# Clone
$ git clone git://github.com/treasure-data/heroku-td-agent.git
$ cd heroku-td-agent
$ rm -fR .git
$ git init
$ git add .
$ git commit -m 'initial commit'

# Create the app & deploy
$ heroku create --stack cedar
$ git push heroku master

# Set your config
$ heroku config:set TREASURE_DATA_API_KEY=you_treasure_data_api_key_value

# Deploy
$ git push heroku master
```

To retrieve your `treasure_data_api_key_value`, see [here](http://docs.treasuredata.com/articles/get-apikey).

## Test

Let's confirm that the log aggregation server has been set up correctly. Please send a GET request to the log server, <http://td-agent-on-heroku.herokuapp.com>, as shown below. This will log the event { "json": "message" } with a debug.sample tag. Note how the JSON-formatted data is passed as a query parameter value.

```
$ curl "http://td-agent-on-heroku.herokuapp.com/debug.sample?json=%7B%22json%22%3A%22message%22%7D"
```

In general, the URL format is

```
http://{YOUR LOG SERVER DOMAIN}/td.{DB_NAME}.{TABLE_NAME}?json={JSON_FORMATTED_DATA}
```

The output will be available on STDOUT.

```
$ heroku logs --tail
```

If this article is incorrect or outdated, or omits critical information, please [let us know](https://github.com/fluent/fluentd-docs-gitbook/issues?state=open). [Fluentd](http://www.fluentd.org/) is a open source project under [Cloud Native Computing Foundation (CNCF)](https://cncf.io/). All components are available under the Apache 2 License.