Fluentd Blog

Fluentd v1.3.0 has been released

Hi users!

We have released v1.3.0. ChangeLog is here.

Output: Change thread execution control

This is internal change of output plugin. This change should fix unexpected error like zlib's DataError.

Output: Change default value of queued_chunks_limit_size

queued_chunks_limit_size is now same as flush_thread_count when queued_chunks_limit_size is not specified.

Before, queued_chunks_limit_size is unlimited. This causes critical situation with smaller flush_interval when the destinaion has a problem. For example, lots of very small queued buffer are created and it causes resource limit error.

Unlimited queued buffer is unexpected behaviour, so we changed this behaviour for safety.

in_syslog: octet counting support

We introduced frame_type parameter. With frame_type octet_count configuration, in_syslog accepts octet counting messages like below.

44 <6>Sep 10 00:00:00 localhost logger: hello!

Major bug fixes

• server: Fix deadlock between on_writable and close in sockets. This fixes fluentd restart issue with in_forward's TLS setting error.

Enjoy logging!

Read More

Fluentd v1.2.5 has been released

Hi users!

We have released v1.2.5. ChangeLog is here. This release fixed in_tail resource leak.

Major bug fixes

• in_tail: Fix log rotation related resource leak. If you use in_tail with log rotation files, we recommend to update fluentd to this version.

Enjoy logging!

Read More

Fluentd v1.2.4 has been released

Hi users!

We have released v1.2.4. ChangeLog is here. This release is for bug fixes.

Major bug fixes

• output: Consider timezone with larger timekey. This fixes unexpected buffer flush with 1d timekey.

• server helper: Fix connection leak by close timing issue

Thanks for submitting bug reports and patches :)

Enjoy logging!

Read More

Fluentd v1.2.2 has been released

Hi users!

We have released v1.2.2. ChangeLog is here. This release is mainly for bug fixes.

filter_parser: Add remove_key_name_field parameter

With reserve_data true, key_name field is kept in the record.

Original record:

{"k1":"v1","log":"{\"k2\":\"v2\"}"}

Parsed result:

{"k1":"v1","log":"{\"k2\":\"v2\"}","k2":"v2"}

But we often don't need original key_name field when parsing is succeeded. You can remove key_name field with remove_key_name_field true. Parsed result is below:

{"k1":"v1","k2":"v2"}

Major bug fixes

• buffer: Wait for all chunks being purged before deleting @queued_num items

Thanks for submitting bug reports and patches :)

Enjoy logging!

Read More

Fluentd v1.2.1 has been released

Hi users!

We have released v1.2.1. ChangeLog is here. This release is mainly for bug fixes.

Counter API: Add wait to Client

This is handy API for error check. Using get, you need to check error response from the response body. wait raises an exception when API call has a problem.

require 'fluent/plugin/filter'

module Fluent
  module Plugin
    class CounterFilter < Filter
      Plugin.register_filter('counter', self)

      helpers :counter

      def start
        super

        @client = counter_client_create(scope: 'test')
        @client.establish('counter')
        # if init call returns an error, exception is raised
        begin
          @client.init(:name => 'num', type: 'numeric', reset_interval: 10).wait
        rescue Fluent::Counter::BaseError
          # process client specific error
        end
      end

      # ...
    end
  end
end

Major bug fixes

• intcp/inudp: Fix sourcehostnamekey to set hostname correctly
• out_file: Temporal fix for broken gzipped files with gzip and append true

Thanks for submitting bug reports and patches :)

Enjoy logging!

Read More