/fluent-plugin-azuresearch

Azure Search Output Plugin for Fluentd

Primary LanguageRuby

Azure Search output plugin for Fluentd

fluent-plugin-azuresearch is a fluent plugin to output to Azure Search

Installation

$ gem install fluent-plugin-azuresearch

Configuration

Azure Search

To use Microsoft Azure Search, you must create an Azure Search service in the Azure Portal. Also you must have an index, persisted storage of documents to which fluent-plugin-azuresearch writes event stream out. Here are instructions:

Fluentd - fluent.conf

<match azuresearch.*>
    type azuresearch
    endpoint   https://AZURE_SEARCH_ACCOUNT.search.windows.net
    api_key    AZURE_SEARCH_API_KEY
    search_index  messages
    column_names id,user_name,message,tag,created_at
    key_names postid,user,content,tag,posttime
</match>
  • endpoint (required) - Azure Search service endpoint URI
  • api_key (required) - Azure Search API key
  • search_index (required) - Azure Search Index name to insert records
  • column_names (required) - Column names in a target Azure search index. Each column needs to be separated by a comma.
  • key_names (optional) - Default:nil. Key names in incomming record to insert. Each key needs to be separated by a comma. ${time} is placeholder for Time.at(time).strftime("%Y-%m-%dT%H:%M:%SZ"), and ${tag} is placeholder for tag. By default, key_names is as same as column_names

Sample Configurations

Case1 - column_names is as same as key_names

Suppose you have the following fluent.conf and azure search index schema:

fluent.conf

<match azuresearch.*>
    type azuresearch
    endpoint   https://yoichidemo.search.windows.net
    api_key    2XX3D2456052A9AD21E54CB03C3ABF6A(dummy)
    search_index  messages
    column_names id,user_name,message,created_at
</match>

Azure Search Schema: messages

{
    "name": "messages",
    "fields": [
        { "name":"id", "type":"Edm.String", "key": true, "searchable": false },
        { "name":"user_name", "type":"Edm.String" },
        { "name":"message", "type":"Edm.String", "filterable":false, "sortable":false, "facetable":false, "analyzer":"en.lucene" },
        { "name":"created_at", "type":"Edm.DateTimeOffset", "facetable":false}
    ]
}

The plugin will write event stream out to Azure Ssearch like this:

Input event stream

{ "id": "1", "user_name": "taylorswift13", "message":"post by taylorswift13", "created_at":"2016-01-29T00:00:00Z" },
{ "id": "2", "user_name": "katyperry", "message":"post by katyperry", "created_at":"2016-01-30T00:00:00Z" },
{ "id": "3", "user_name": "ladygaga", "message":"post by ladygaga", "created_at":"2016-01-31T00:00:00Z" }

Search results

"value": [
    { "@search.score": 1, "id": "1", "user_name": "taylorswift13", "message": "post by taylorswift13", "created_at": "2016-01-29T00:00:00Z" },
    { "@search.score": 1, "id": "2", "user_name": "katyperry", "message": "post by katyperry", "created_at": "2016-01-30T00:00:00Z" },
    { "@search.score": 1, "id": "3", "user_name": "ladygaga", "message": "post by ladygaga", "created_at": "2016-01-31T00:00:00Z" }
]

Case2 - column_names is NOT as same as key_names

Suppose you have the following fluent.conf and azure search index schema:

fluent.conf

<match azuresearch.*>
    type azuresearch
    endpoint   https://yoichidemo.search.windows.net
    api_key    2XX3D2456052A9AD21E54CB03C3ABF6A(dummy)
    search_index  messages
    column_names id,user_name,message,created_at
    key_names postid,user,content,posttime
</match>

Azure Search Schema: messages

{
    "name": "messages",
    "fields": [
        { "name":"id", "type":"Edm.String", "key": true, "searchable": false },
        { "name":"user_name", "type":"Edm.String" },
        { "name":"message", "type":"Edm.String", "filterable":false, "sortable":false, "facetable":false, "analyzer":"en.lucene" },
        { "name":"created_at", "type":"Edm.DateTimeOffset", "facetable":false}
    ]
}

The plugin will write event stream out to Azure Ssearch like this:

Input event stream

{ "postid": "1", "user": "taylorswift13", "content":"post by taylorswift13", "posttime":"2016-01-29T00:00:00Z" },
{ "postid": "2", "user": "katyperry", "content":"post by katyperry", "posttime":"2016-01-30T00:00:00Z" },
{ "postid": "3", "user": "ladygaga", "content":"post by ladygaga", "posttime":"2016-01-31T00:00:00Z" }

Search results

"value": [
    { "@search.score": 1, "id": "1", "user_name": "taylorswift13", "message": "post by taylorswift13", "created_at": "2016-01-29T00:00:00Z" },
    { "@search.score": 1, "id": "2", "user_name": "katyperry", "message": "post by katyperry", "created_at": "2016-01-30T00:00:00Z" },
    { "@search.score": 1, "id": "3", "user_name": "ladygaga", "message": "post by ladygaga", "created_at": "2016-01-31T00:00:00Z" }
]

Case3 - column_names is NOT as same as key_names, Plus, key_names includes ${time} and ${tag}

fluent.conf

<match azuresearch.*>
    type azuresearch
    endpoint   https://yoichidemo.search.windows.net
    api_key    2XX3D2456052A9AD21E54CB03C3ABF6A(dummy)
    search_index  messages
    column_names id,user_name,message,tag,created_at
    key_names postid,user,content,${tag},${time}
</match>

Azure Search Schema: messages

{
    "name": "messages",
    "fields": [
        { "name":"id", "type":"Edm.String", "key": true, "searchable": false },
        { "name":"user_name", "type":"Edm.String" },
        { "name":"message", "type":"Edm.String", "filterable":false, "sortable":false, "facetable":false, "analyzer":"en.lucene" },
        { "name":"created_at", "type":"Edm.DateTimeOffset", "facetable":false}
    ]
}

The plugin will write event stream out to Azure Ssearch like this:

Input event stream

{ "id": "1", "user_name": "taylorswift13", "message":"post by taylorswift13" },
{ "id": "2", "user_name": "katyperry", "message":"post by katyperry" },
{ "id": "3", "user_name": "ladygaga", "message":"post by ladygaga" }

Search results

"value": [
    { "@search.score": 1, "id": "1", "user_name": "taylorswift13", "message": "post by taylorswift13", "tag": "azuresearch.msg", "created_at": "2016-01-31T21:03:41Z" },
    { "@search.score": 1, "id": "2", "user_name": "katyperry", "message": "post by katyperry", "tag": "azuresearch.msg", "created_at": "2016-01-31T21:03:41Z" },
    { "@search.score": 1, "id": "3", "user_name": "ladygaga", "message": "post by ladygaga", "tag": "azuresearch.msg", "created_at": "2016-01-31T21:03:41Z" }
]

[note] the value of created_at above is the time when fluentd actually recieves its corresponding input event.

Tests

Running test code

$ git clone https://github.com/yokawasa/fluent-plugin-azuresearch.git
$ cd fluent-plugin-azuresearch

# edit CONFIG params of test/plugin/test_azuresearch.rb 
$ vi test/plugin/test_azuresearch.rb

# run test 
$ rake test

Creating package, running and testing locally

$ rake build
$ rake install:local
 
# running fluentd with your fluent.conf
$ fluentd -c fluent.conf -vv &
 
# send test input event to test plugin using fluent-cat
$ echo ' { "postid": "100", "user": "ladygaga", "content":"post by ladygaga"}' | fluent-cat azuresearch.msg

Please don't forget that you need forward input configuration to receive the message from fluent-cat

<source>
    type forward
</source>

TODOs

  • Input validation for Azure Search - check total size of columns to add

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/yokawasa/fluent-plugin-azuresearch.

Copyright

CopyrightCopyright (c) 2016- Yoichi Kawasaki
LicenseApache License, Version 2.0