Fluentd JDK log handler implementation

This project provides JDK logging handler implementation that logs to fluentd daemon. The handler uses fluency library, which sends logging data to an in_forward input plugin.

The handler requires JDK 8 or above to run.

Details on JDK logging can be found here.

JavaDocs for the implementation (latest version) can be found here

This code is available at Maven Central

<dependency>
    <groupId>codes.vps</groupId>
    <artifactId>fluentd-jdk-handler</artifactId>
    <version>0.5</version>
</dependency>

The project also produced a jar with all dependencies which is more suitable for DevOps integration. You can use the following Maven command to retrieve the jar:

$ mvn org.apache.maven.plugins:maven-dependency-plugin:3.1.2:get\
 -Dartifact=codes.vps:fluentd-jdk-handler:0.5:jar:jar-with-dependencies

The JAR will then be downloaded into (unless you've changed your local Maven repository location) to ~/.m2/repository/codes/vps/fluentd-jdk-handler/0.5/fluentd-jdk-handler-0.5-jar-with-dependencies.jar

This can be used in scripts and so on.

When invoked through API, configuration information is to be provided in the Builder object to the handler, see example

When configuring through a logging framework property files, the handler shall be configured with its classname, FluentdHandler, and the following properties are available:

• FluentdHandler.tag_prefix, default is an empty string
  Specifies tag prefix for all messages sent through the corresponding fluentd logger.
• FluentdHandler.host
  Specifies host name to send messages to, default is 127.0.0.1
• FluentdHandler.port
  Specifies port to send messages to, default is 24224
• FluentdHandler.format
  Specifies formatting string (see Formatting) below. Default is tag"";message"${level10n} [${tid}] ${class}.${method} ${l10n}";stack"${trace}".
• Fluency configuration options; please see fluency for the additional documentation on those.
  • FluentdHandler.sender_max_retry_count
    Maximum retry count, default is 7
  • FluentdHandler.sender_base_retry_interval_millis
    Initial retry interval, in milliseconds, default is 400
  • FluentdHandler.sender_max_retry_interval_millis
    Maximum retry interval, in milliseconds, default is 30000
  • FluentdHandler.ack_response_mode
    Request acknowledgement for packets sent to fluentd, default is false
  • FluentdHandler.ssl_enabled
    Specifies whether SSL connection should be used, default is false
  • FluentdHandler.connection_timeout_milli
    Specified connection timeout, in milliseconds, default is 5000
  • FluentdHandler.read_timeout_milli
    Specified socket read timeout, in milliseconds, default is 5000
  • FluentdHandler.max_buffer_size
    Maximum buffer size to use, in bytes, default is 536870912
  • FluentdHandler.buffer_chunk_initial_size
    Initial buffer chunk size, default is 1048576
  • FluentdHandler.buffer_chunk_retention_size
    Threshold chunk buffer size to flush, default is 4194304
  • FluentdHandler.buffer_chunk_retention_time_millis
    Threshold time to flush the buffer, in milliseconds, default is 1000
  • FluentdHandler.flush_attempt_interval_millis
    Flush attempt interval, in milliseconds, default is 600
  • FluentdHandler.file_backup_dir
    Directory where the logging message shall be backed up in, default is not set.
  • FluentdHandler.wait_until_buffer_flushed
    Time to wait until the buffer is flushed, in seconds, default is 60
  • FluentdHandler.wait_until_flusher_terminated
    Time to wait until the flusher is terminated, in seconds, default is 60
  • FluentdHandler.jvm_head_buffer_mode
    Specified whether to enable heap buffer memory (true) or off-heap buffer memory (false), default is false.

fluency accepts the following parameters when logging a single message:

• tag
• timestamp
• map with arbitrary key/value pairs (values are objects)

The formatter process takes a LogRecord object and converts it into the input suitable for fluency. Format definition specifies a series of statements that indicate which keys in that arbitrary map should be populated with which values. Then, keys with names "$tag" and "$timestamp" are treated specially, they are removed from the map, and fed as tag and timestamp parameters directly into fluency.

If tag ends up being not specified, it is populated from logger name value of the log record. If timestamp ends up being not specified, then it is populated from millis property of the log record.

Formatter string is defined as follows:

• format := item [ ';' item ... ]
• item := field '"' format '"' [ type ]
• field := <literal map field name>
• type := 's' | 'n' | 'b' (string, number, or boolean)
• format := <format string to generate value>

The literals in format string will be copied (after escaping) to the output as is. variables can, however, be referenced using ${...}, e.g. ${level}. When referencing millis - additional date format, after ,, can be provided, in this case the value will be passed through a date formatter: SimpleDateFormatter. For example: date"${millis,yyyy-MM-dd'T'HH:mm:ss.SSSZ}

OS environment variables are also supported and can be referenced using $[...], e.g. $[PATH]. If an environment variable is not set, the reference is replaced with an empty string.

Any character can be escaped from current level of processing by specifying backslash (\ ) character in front of it. To insert backslash itself, simply escape it (\\ parses as \ ). Escaping must be done on multiple levels if need to escape nested structures, i.e. the text is "unescaped" when:

• items are extracted
• field, type, format are extracted
• contents of format are extracted

List of variables that can be referenced (based on LogRecord class parameters):

• level - log level
• level10n - localized log level
• sequence - log sequence
• class - source class
• method - source method
• message - original message value as is;
• l10n - localized message, message+parameters will be passed through l10n
• params - localization parameters (printed as comma-separated string representations)
• millis - timestamp
• logger - name of the logger
• tid - thread ID
• trace - entire stack trace of an attached exception, if any, or an empty string

Example format: logger"${logger}";level"${level}";$timestamp"${millis}n";message"${l10n};path"$[PATH]"