http

The in_http Input plugin allows you to send events through HTTP requests. Using this plugin, you can trivially launch a REST endpoint to gather data.

Configuration

Here is a sample configuration:

<source>
  @type http
  port 9880
  bind 0.0.0.0
  body_size_limit 32m
  keepalive_timeout 10s
</source>

For the full list of the configurable options, see the Parameters section.

Basic Usage

Here is a simple example to post a record using curl, which uses the default Content-Type application/x-www-form-urlencoded

# Post a record with the tag "app.log"
$ curl -X POST -d 'json={"foo":"bar"}' http://localhost:9880/app.log

For more details regarding the message body syntax and Content-Type see How to use HTTP Content-Type Header

By default, timestamps are assigned to each record on arrival. You can override the timestamp using the time parameter:

# Overwrite the timestamp to 2018-02-16 04:40:37.3137116
$ curl -X POST -d 'json={"foo":"bar"}' \
  http://localhost:9880/test.tag?time=1518756037.3137116

Here is another example in JavaScript:

// Post a record using XMLHttpRequest
var form = new FormData();
form.set('json', JSON.stringify({"foo": "bar"}));

var req = new XMLHttpRequest();
req.open('POST', 'http://localhost:9880/debug.log');
req.send(form);

For more advanced usage, please read the Tips and Tricks section.

Parameters

See Common Parameters.

@type (required)

The value must be http.

port

typedefaultversion

integer

9880

0.14.0

The port to listen to.

bind

typedefaultversion

string

0.0.0.0 (all addresses)

0.14.0

The bind address to listen to.

body_size_limit

typedefaultversion

size

32MB

0.14.0

The size limit of the POSTed element.

keepalive_timeout

typedefaultversion

size

10 (seconds)

0.14.0

The timeout limit for keeping the connection alive.

add_http_headers

typedefaultversion

bool

false

0.14.0

Adds HTTP_ prefix headers to the record.

add_remote_addr

typedefaultversion

bool

false

0.14.0

Adds REMOTE_ADDR field to the record. The value of REMOTE_ADDR is the client's address.

If your system set multiple X-Forwarded-For headers in the request, in_http uses the first one. For example:

X-Forwarded-For: host1, host2
X-Forwarded-For: host3

If the above multiple headers are sent, the value of REMOTE_ADDR will be host1.

cors_allow_origins

typedefaultversion

array

nil(disabled)

0.14.0

Whitelist domains for CORS.

If you set ["domain1", "domain2"] to cors_allow_origins, in_http returns 403 to access from other domains. Since Fluentd v1.2.6, you can use a wildcard character * to allow requests from any origins.

Example:

<source>
  @type http
  port 9880
  cors_allow_origins ["*"]
</source>

cors_allow_credentials

typedefaultversion

bool

false

1.14.0

Add Access-Control-Allow-Credentials header. It's needed when a request's credentials mode is include. An example of use case is using Beacon API, its request mode is always include.

respond_with_empty_img

typedefaultversion

bool

false

0.12.0

Responds with an empty GIF image of 1x1 pixel (rather than an empty string).

use_204_response

typedefaultversion

bool

false

v1.8.0

Respond status code with 204. This option will be deprecated at v2 because fluentd v2 will respond 204 as default.

<transport> Section

typedefaultavailable valuesversion

enum

tcp

tcp, tls

1.5.0

This section is for setting TLS transport or some general transport configurations.

General configuration

linger_timeout

typedefaultavailable transport typeversion

integer

0

tcp, tls

1.14.6

The timeout (seconds) to set SO_LINGER.

The default value 0 is to send RST rather than FIN to avoid lots of connections sitting in TIME_WAIT on closing on non-Windows.

You can set positive value to send FIN on closing on non-Windows.

(On Windows, Fluentd sends FIN when linger_timeout is 0 too).

<transport tcp>
  linger_timeout 1
</transport>

TLS configuration

<transport tls>
  cert_path /path/to/fluentd.crt
  # other parameters
</transport>

See How to Enable TLS Encryption section for how to use and see Configuration Example for all supported parameters.

Without <transport tls>, in_http uses HTTP.

<parse> directive

Use the parser plugin to parse the incoming data. See also Handle other formats using parser plugins section.

format (deprecated)

Deprecated parameter. Use <parse> directive instead.

Tips and Tricks

How to send data in MessagePack format?

You can post data in MessagePack format by adding the msgpack= prefix:

# Send data in msgpack format
$ msgpack=`echo -e "\x81\xa3foo\xa3bar"`
$ curl -X POST -d "msgpack=$msgpack" http://localhost:9880/app.log

How to use HTTP Content-Type header?

in_http plugin recognizes HTTP Content-Type header in the incoming requests.

By default curl uses -H "Content-Type: application/x-www-form-urlencoded", which allows the use of the syntax json= and msgpack= as seen on the previous examples.

However, you can send a JSON payload without the json= prefix by setting the content type application/json:

$ curl -X POST -d '{"foo":"bar"}' -H 'Content-Type: application/json' \
  http://localhost:9880/app.log

To use MessagePack, set the content type to application/msgpack:

$ msgpack=`echo -e "\x81\xa3foo\xa3bar"`
$ curl -X POST -d "$msgpack" -H 'Content-Type: application/msgpack' \
  http://localhost:9880/app.log

Handle Other Formats using Parser Plugins

You can handle various input formats by using the <parse> directive. For example, add the following settings to the configuration file:

<source>
  @type http
  port 9880
  <parse>
    @type regexp
    expression /^(?<field1>\d+):(?<field2>\w+)$/
  </parse>
</source>

Now you can post custom-format records like this:

# This will be parsed into {"field1":"123456","field2":"awesome"}
$ curl -X POST -d '123456:awesome' http://localhost:9880/app.log

Many other formats (e.g. csv/syslog/nginx) are also supported. For the full list of supported formats, see Parser Plugin Overview.

NOTE: Some parser plugins do not support the batch mode. So, if you want to use bulk insertion for handling a large data set, please consider keeping the default JSON (or MessagePack) format or write batch mode supported parser (return array object).

Enhance Performance

Handle Large Data with Batch Mode

You can post multiple records with a single request by packing data into a JSON/MessagePack array:

# Send multiple events as a JSON array
$ curl -X POST -d 'json=[{"foo":"bar"},{"abc":"def"},{"xyz":"123"}]' \
  http://localhost:9880/app.log

This significantly improves the throughput since it reduces the number of HTTP requests. Here is a simple benchmark on MacBook Pro with Ruby 2.3:

jsonmsgpackmsgpack array(10 items)

2100 events/sec

2400 events/sec

10000 events/sec

Tested configuration and Ruby script are here.

Use Compression to Reduce Bandwidth Overhead

Since v1.2.3, Fluentd can handle gzip-compressed payloads. To enable this feature, you need to add the Content-Encoding header to your requests.

# Send gzip-compressed payload
$ echo 'json={"foo":"bar"}' | gzip > json.gz
$ curl --data-binary @json.gz -H "Content-Encoding: gzip" \
  http://localhost:9880/app.log

You do not need any configuration to enable this feature.

Multi-process Environment

If you use this plugin under the multi-process environment, the port will be shared.

<system>
  workers 3
</system>
<source>
  @type http
  port 9880
</source>

With this configuration, three (3) workers share 9880 port. No need for an additional port. Incoming data will be routed to three (3) workers automatically.

Troubleshooting

Why in_http splits the messages from my log when using json= syntax ?

This happens when using the content type application/x-www-form-urlencoded. It is a limitation of the HTTP spec, you either need to encode the message or you can use the content type application/json.

For example:


# bad
curl -X POST --location "http://localhost:9880/app.loge" \
        -d "json=[{\"log\":\"message ; remaining message\"}]"
# good
curl -X POST --location "http://localhost:9880/app.log" \
        -H "Content-Type: application/json" \
        -d "{\"log\":\"message ; remaining message\"}"

Why in_http removes '+' from my log?

This is HTTP spec, not fluentd problem. You need to encode your payload properly or use multipart request. Here is a Ruby example:

# Good
URI.encode_www_form({json: {"message" => "foo+bar"}.to_json})

# Bad
"json=#{"message" => "foo+bar"}.to_json}"

curl command example:

# Good
curl -X POST -H 'Content-Type: multipart/form-data' -F 'json={"message":"foo+bar"}' http://localhost:9880/app.log

# Bad
curl -X POST -F 'json={"message":"foo+bar"}' http://localhost:9880/app.log

Learn More

Tips

How to Enable TLS Encryption?

Since v1.5.0, in_http support TLS transport. Here is a configuration example with HTTPS client:

<source>
  @type http
  bind 0.0.0.0
  <transport tls>
    ca_path /etc/pki/ca.pem
    cert_path /etc/pki/cert.pem
    private_key_path /etc/pki/key.pem
    private_key_passphrase PASSPHRASE
  </transport>
</source>
  • https client

require 'net/http'
require 'net/https'
require 'msgpack'

record = { 'msgpack' => { 'k' => 'hello', 'k1' => 1234}.to_msgpack }

def post(path, params)
  http = Net::HTTP.new('127.0.0.1', 9880)
  http.use_ssl = true
  http.verify_mode = OpenSSL::SSL::VERIFY_NONE
  req = Net::HTTP::Post.new(path, {})
  req.set_form_data(params)
  http.request(req)
end

puts post("/test.http?time=#{Time.now.to_i}", record).body

How to Enable TLS Mutual Authentication?

Fluentd supports TLS mutual authentication (i.e. client certificate auth). If you want to use this feature, please set the client_cert_auth and ca_path options like this:

<source>
  @type http
  <transport tls>
    ...
    client_cert_auth true
    ca_path /path/to/ca/cert
  </transport>
</source>

When this feature is enabled, Fluentd will check all the incoming requests for a client certificate signed by the trusted CA. Requests with an invalid client certificate will fail.

If this article is incorrect or outdated, or omits critical information, please let us know. Fluentd is an open-source project under Cloud Native Computing Foundation (CNCF). All components are available under the Apache 2 License.

Last updated