= Amazon S3 output plugin for {Fluentd}[http://github.com/fluent/fluentd]

== Overview

*s3* output plugin buffers event logs in local file and upload it to S3 periodically.

This plugin splits files exactly by using the time of event logs (not the time when the logs are received). For example, a log '2011-01-02 message B' is reached, and then another log '2011-01-03 message B' is reached in this order, the former one is stored in "20110102.gz" file, and latter one in "20110103.gz" file.


== Installation

Simply use RubyGems:

    gem install fluent-plugin-s3

== Configuration

    <match pattern>
      type s3

      aws_key_id YOUR_AWS_KEY_ID
      aws_sec_key YOUR_AWS_SECRET/KEY
      s3_bucket YOUR_S3_BUCKET_NAME
      s3_region ap-northeast-1
      s3_object_key_format %{path}%{time_slice}_%{index}.%{file_extension}
      path logs/
      buffer_path /var/log/fluent/s3

      time_slice_format %Y%m%d-%H
      time_slice_wait 10m
      utc
    </match>

[aws_key_id] AWS access key id. This parameter is required when your agent is not running on EC2 instance with an IAM Role.

[aws_sec_key] AWS secret key. This parameter is required when your agent is not running on EC2 instance with an IAM Role.

[aws_iam_retries] The number of attempts to make (with exponential backoff) when loading instance profile credentials from the EC2 metadata service using an IAM role. Defaults to 5 retries.

[s3_bucket (required)] S3 bucket name.

[s3_region] s3 region name. For example, US West (Oregon) Region is "us-west-2". The full list of regions are available here. > http://docs.aws.amazon.com/general/latest/gr/rande.html#s3_region. We recommend using `s3_region` instead of `s3_endpoint`.

[s3_endpoint] endpoint for S3 compatible services. For example, Riak CS based storage or something. This option doesn't work on S3, use `s3_region` instead.

[s3_object_key_format] The format of S3 object keys. You can use several built-in variables:

- %{path}
- %{time_slice}
- %{index}
- %{file_extension}

to decide keys dynamically.

%{path} is exactly the value of *path* configured in the configuration file. E.g., "logs/" in the example configuration above.
%{time_slice} is the time-slice in text that are formatted with *time_slice_format*.
%{index} is the sequential number starts from 0, increments when multiple files are uploaded to S3 in the same time slice.
%{file_extention} is always "gz" for now.

The default format is "%{path}%{time_slice}_%{index}.%{file_extension}".

For instance, using the example configuration above, actual object keys on S3 will be something like:

    "logs/20130111-22_0.gz"
    "logs/20130111-23_0.gz"
    "logs/20130111-23_1.gz"
    "logs/20130112-00_0.gz"

With the configuration:

    s3_object_key_format %{path}/events/ts=%{time_slice}/events_%{index}.%{file_extension}
    path log
    time_slice_format %Y%m%d-%H

You get:

    "log/events/ts=20130111-22/events_0.gz"
    "log/events/ts=20130111-23/events_0.gz"
    "log/events/ts=20130111-23/events_1.gz"
    "log/events/ts=20130112-00/events_0.gz"

The {fluent-mixin-config-placeholders}[https://github.com/tagomoris/fluent-mixin-config-placeholders] mixin is also incorporated, so additional variables such as %{hostname}, %{uuid}, etc. can be used in the s3_object_key_format. This could prove useful in preventing filename conflicts when writing from multiple servers.

    s3_object_key_format %{path}/events/ts=%{time_slice}/events_%{index}-%{hostname}.%{file_extension}

[store_as] archive format on S3. You can use serveral format:

- gzip (default)
- json
- text
- lzo (Need lzop command)

See 'Use your compression algorithm' section for adding another format.

[format] Change one line format in the S3 object. Supported formats are "out_file", "json", "ltsv" and "single_value".

- out_file (default).

    time\ttag\t{..json1..}
    time\ttag\t{..json2..}
    ...

- json

    {..json1..}
    {..json2..}
    ...

At this format, "time" and "tag" are omitted.
But you can set these information to the record by setting "include_tag_key" / "tag_key" and "include_time_key" / "time_key" option.
If you set following configuration in S3 output:

    format json
    include_time_key true
    time_key log_time # default is time

then the record has log_time field.

    {"log_time":"time string",...}

- ltsv

    key1:value1\tkey2:value2
    key1:value1\tkey2:value2
    ...

"ltsv" format also accepts "include_xxx" related options. See "json" section.

- single_value

Use specified value instead of entire recode. If you get '{"message":"my log"}', then contents are

    my log1
    my log2
    ...

You can change key name by "message_key" option.

[auto_create_bucket] Create S3 bucket if it does not exists. Default is true.

[check_apikey_on_start] Check AWS key on start. Default is true.

[proxy_uri] uri of proxy environment.

[path] path prefix of the files on S3. Default is "" (no prefix).

[buffer_path (required)] path prefix of the files to buffer logs.

[time_slice_format] Format of the time used as the file name. Default is '%Y%m%d'. Use '%Y%m%d%H' to split files hourly.

[time_slice_wait] The time to wait old logs. Default is 10 minutes. Specify larger value if old logs may reache.

[utc] Use UTC instead of local time.

[reduced_redundancy] Use S3 reduced redundancy storage for 33% cheaper pricing. Default is false.

== IAM Policy

The following is an example for a minimal IAM policy needed to write to an s3 bucket (matches my-s3bucket/logs, my-s3bucket-test, etc.).

    { "Statement": [
     { "Effect":"Allow",
       "Action":"s3:*",
       "Resource":"arn:aws:s3:::my-s3bucket*"
      } ]
    }

Note that the bucket must already exist and *auto_create_bucket* has no effect in this case.

Refer to the {AWS documentation}[http://docs.aws.amazon.com/IAM/latest/UserGuide/ExampleIAMPolicies.html] for example policies.

Using {IAM roles}[http://docs.aws.amazon.com/IAM/latest/UserGuide/WorkingWithRoles.html] with a properly configured IAM policy are preferred over embedding access keys on EC2 instances.

== Use your compression algorithm

s3 plugin has plugabble compression mechanizm like Fleuntd\'s input / output plugin.
If you set 'store_as xxx', s3 plugin searches `fluent/plugin/s3_compressor_xxx.rb`.
You can define your compression with 'S3Output::Compressor' class. Compressor API is here:

    module Fluent
      class S3Output
        class XXXCompressor < Compressor
          S3Output.register_compressor('xxx', self)

          # Used to file extension
          def ext
            'xxx'
          end

          # Used to file content type
          def content_type
            'application/x-xxx'
          end

          # chunk is buffer chunk. tmp is destination file for upload
          def compress(chunk, tmp)
            # call command or something
          end
        end
      end
    end

See bundled Compressor classes for more detail.

== Website, license, et. al.

Web site::  http://fluentd.org/
Documents:: http://docs.fluentd.org/
Source repository:: http://github.com/fluent
Discussion:: http://groups.google.com/group/fluentd
Author:: Sadayuki Furuhashi
Copyright:: (c) 2011 FURUHASHI Sadayuki
License:: Apache License, Version 2.0