$fluent-gem install fluent-plugin-redis-multi-type-counter

<match redis_counter.**>
  type redis_multi_type_counter

  host localhost
  port 6379

  # database number is optional.
  db_number 0        # 0 is default

  # match condition
  # this pattern matches {"status": "200", "url": "http://foo.example.com"} and then,
  # increment Redis key "foo-status2xx" by calling Redis::incrby( "foo-status2xx", 1 ).
  <pattern>
    match_status ^2[0-9][0-9]$        # matches with {"status": "200", ...
    match_url ^http:\/\/foo\.         # matches with {"url": "http://foo.example.com", ...
    count_key foo-status2xx           # key-name for Redis
    count_value 1                     # count-up amount(default: 1, negative value is allowed)
  </pattern>

  # time-dependent redis keyname
  # for example, "foo-status2xx-%Y-%m-%d" will be formatted to "foo-status2xx-2012-06-21".
  # rules for placeholder(%Y, etc.) is similar to out_file plugin.
  <pattern>
    match_status ^2[0-9][0-9]$
    count_key_format foo-statux2xx-%d
    localtime                         # time-zone(default: localtime, it can be "utc" or "localtime")
  </pattern>

  # you can also use values from the matched JSON record in the key name, by using the 
  # syntax foo%_{key1}-%_{key2}, where key1 and key2 are keys in the JSON record that 
  # was received by fluentd.
  # for example, "customer:%_{customer_id}:status2xx" will be formatted to 
  # "customer:123:status2xx" if the JSON record contains a key named "customer_id" 
  # with value 123, like so: {"status": 200, "customer_id": 123 ... }.
  # these can be combined with the time formatting options in the previous example.
  <pattern>
    match_status ^2[0-9][0-9]$
    count_key_format customer:%_{customer_id}:status2xx-%Y-%m-%d
  </pattern>

  # you can just store value to list (not counting on, it's something awkward) by store_list to true
  # note that you cannot use store_list with hash or zset
  <pattern>
    match_status ^2[0-9][0-9]$
    count_key_format customer:%_{customer_id}:status2xx-%Y-%m-%d
    store_list true
  </pattern>

  # you can also sum up key in hash, by configuring count_hash_key_format
  # syntax is same to count_key_format
  # for example, {"custom_id": 123, "date": "20131219" ...}.
  # HINCRBY item_count:123 20131219 1
  <pattern>
    count_key_format item_count:%_{item_id}
    count_hash_key_format %_{date}
  <pattern>

  # you can also sum up key in sorted set(zset), by configuring count_zset_key_format
  # syntax and usage is same to count_hash_key_format
  # for example, {"custom_id": 123, "date": "20131219" ...}.
  # ZINCRBY item_count:123 1 20131219
  <pattern>
    count_key_format item_count:%_{item_id}
    count_zset_key_format %_{date}
  <pattern>

  # you can also sum up specified key with count_value_key option.
  # for example, {"count": 321, "customer_id": 123 ... }.
  # INCRBY item_count:123 321.
  <pattern>
    count_key_format item_count:%_{item_id}
    count_value_key count
  </pattern>
</match>

<source>
  type forward
</source>
<match debug.**>
  type redis_multi_type_counter
  host localhost
  port 6379
  db_number 0
  <pattern>
    match_status ^2[0-9][0-9]$
    match_url ^http:\/\/foo\.
    count_key foo
  </pattern>
</match>

$redis-server 2>&1 >/dev/null &
[1] 879
$echo del foo | redis-cli -h localhost -p 6379 -n 0
(integer) 0
$fluentd -c ./fluent.conf 2>&1 >/dev/null &
[2] 889
$echo {\"status\": \"200\", \"url\": \"http://foo.example.com\"} | fluent-cat debug
$echo {\"status\": \"500\", \"url\": \"http://foo.example.com\"} | fluent-cat debug
$kill -s HUP 889
$echo get foo | redis-cli -h localhost -p 6379 -n 0
"1"