= MongoDB plugin for Fluentd

fluent-plugin-mongo provides input and output plugins for {Fluentd}[http://fluentd.org/].

= Installation

== Gems

The gem is hosted at {Rubygems.org}[http://rubygems.org]. You can install the gem as follows:

    $ fluent-gem install fluent-plugin-mongo

= Plugins

== Output plugin

=== mongo

Store Fluentd event to MongoDB database.

==== Configuration

Use _mongo_ type in match.

    <match mongo.**>
      type mongo
      database fluent
      collection test

      # Following attibutes are optional
      host fluenter
      port 10000

      # Set 'capped' if you want to use capped collection
      capped
      capped_size 100m

      # Set 'user' and 'password' for authentication
      user handa
      password shinobu

      # Other buffer configurations here
    </match>

=== mongo(tag mapped mode)

Tag mapped to MongoDB collection automatically.

==== Configuration

Use _tag_mapped_ parameter in match of _mongo_ type.

If tag name is "foo.bar", auto create collection "foo.bar" and insert data.

    <match forward.*>
      type mongo
      database fluent

      # Set 'tag_mapped' if you want to use tag mapped mode.
      tag_mapped

      # If tag is "forward.foo.bar", then prefix "forward." is removed.
      # Collection name to insert is "foo.bar".
      remove_tag_prefix forward.

      # This configuration is used if tag not found. Default is 'untagged'.
      collection misc

      # Other configurations here
    </match>

=== mongo_replset

Replica Set version of mongo.

==== Configuration

Use _mongo_replset_ type in match.

    <match mongo.**>
      type mongo_replset
      database fluent
      collection logs

      # each node separated by ','
      nodes localhost:27017,localhost:27018,localhost:27019

      # num_retries is threshold at failover, default is 60.
      # If retry count reached this threshold, mongo plugin raises an exception.
      num_retries 30

      # following optional parameters passed to ReplSetConnection of mongo-ruby-driver.
      # See mongo-ruby-driver docs for more detail.
      #name replset_name
      #read secondary
      #refresh_mode sync
      #refresh_interval 60
    </match>

=== mongo_backup

Store Fluentd event to local capped collection for backup.

==== Configuration

Use _mongo_backup_ type in match. _mongo_backup_ alwalys use capped collection.

    <match ...>
      type mongo_backup
      capped_size 100m

      <store>
        type tcp
        host 192.168.0.13
        ...
      </store>
    </match>

== Input plugin

=== mongo_tail

Tail capped collection to input data.

==== Configuration

Use _mongo_tail_ type in source.

    <source>
      type mongo_tail
      database fluent
      collection capped_log

      tag app.mongo_log

      # Convert 'time'(BSON's time) to fluent time(Unix time).
      time_key time

      # You can store last ObjectId to tail over server's shutdown
      id_store_file /Users/repeatedly/devel/fluent-plugin-mongo/last_id
    </source>

= NOTE

== Broken data as a BSON

Fluentd event sometimes has an invalid record as a BSON.
In such case, Mongo plugin marshals an invalid record using Marshal.dump
and re-inserts its to same collection as a binary.

If passed following invalid record:

    {"key1": "invalid value", "key2": "valid value", "time": ISODate("2012-01-15T21:09:53Z") }

then Mongo plugin converts this record to following format:

    {"__broken_data": BinData(0, Marshal.dump result of {"key1": "invalid value", "key2": "valid value"}), "time": ISODate("2012-01-15T21:09:53Z") }

Mongo-Ruby-Driver cannot detect an invalid attribute,
so Mongo plugin marshals all attributes excluding Fluentd keys("tag_key" and "time_key").

You can deserialize broken data using Mongo and Marshal.load. Sample code is below:

    # _collection_ is an instance of Mongo::Collection
    collection.find({'__broken_data' => {'$exists' => true}}).each do |doc|
      p Marshal.load(doc['__broken_data'].to_s) #=> {"key1": "invalid value", "key2": "valid value"}
    end

If you want to ignore an invalid record, set _true_ to _ignore_invalid_record_ parameter in match.

    <match forward.*>
      ...

      # ignore invalid documents at write operation
      ignore_invalid_record true

      ...
    </match>

== Buffer size limitation

Mongo plugin has the limitation of buffer size.
Because MongoDB and mongo-ruby-driver checks the total object size at each insertion.
If total object size gets over the size limitation, then
MongoDB returns error or mongo-ruby-driver raises an exception.

So, Mongo plugin resets _buffer_chunk_limit_ if configurated value is larger than above limitation:
- Before v1.8, max of _buffer_chunk_limit_ is 2MB
- After v1.8, max of _buffer_chunk_limit_ is 10MB

== Disable collection check

Mongo plugin checks a collection's configuration to prevent unexpected insertion to existing collection.
In tag mapped mode, Mongo plugin accesses many collections. In this case, collection checking is not usable
because current configuration format cannot write multiple values in one parameter.
So, if you disable this checking, put ```disable_collection_check true``` in match.

    <match forward.*>
      ...

      disable_collection_check true

      ...
    </match>

= Tool

You can tail mongo capped collection.

    $ mongo-tail -f

= TODO

== More configuration

- Multi process
- etc

= Copyright

Copyright:: Copyright (c) 2011- Masahiro Nakagawa
License::   Apache License, Version 2.0