Sergey Gladkov > Kafka-0.8008 > Kafka

Download:
Kafka-0.8008.tar.gz

Dependencies

Annotate this POD

CPAN RT

New  2
Open  0
View/Report Bugs
Module Version: 0.8008   Source   Latest Release: Kafka-0.8008_1

NAME ^

Kafka - Apache Kafka interface for Perl.

VERSION ^

This documentation refers to Kafka package version 0.8008 .

SYNOPSIS ^

    use 5.010;
    use strict;
    use warnings;

    use Scalar::Util qw(
        blessed
    );
    use Try::Tiny;

    use Kafka qw(
        $BITS64
    );
    use Kafka::Connection;

    # A simple example of Kafka usage

    # common information
    say 'This is Kafka package ', $Kafka::VERSION;
    say 'You have a ', $BITS64 ? '64' : '32', ' bit system';

    my ( $connection, $producer, $consumer );
    try {

        #-- Connect to local cluster
        $connection = Kafka::Connection->new( host => 'localhost' );
        #-- Producer
        $producer = Kafka::Producer->new( Connection => $connection );
        #-- Consumer
        $consumer = Kafka::Consumer->new( Connection  => $connection );

    } catch {
        if ( blessed( $_ ) && $_->isa( 'Kafka::Exception' ) ) {
            warn 'Error: (', $_->code, ') ',  $_->message, "\n";
            exit;
        } else {
            die $_;
        }
    };

    # cleaning up
    undef $consumer;
    undef $producer;
    undef $connection;

    # another brief code example of the Kafka package
    # is provided in the "An Example" section.

ABSTRACT ^

The Kafka package is a set of Perl modules which provides a simple and consistent application programming interface (API) to Apache Kafka 0.8, a high-throughput distributed messaging system.

DESCRIPTION ^

The user modules in this package provide an object oriented API. The IO agents, requests sent, and responses received from the Apache Kafka or mock servers are all represented by objects. This makes a simple and powerful interface to these services.

The main features of the package are:

APACHE KAFKA'S STYLE COMMUNICATION ^

The Kafka package is based on Kafka's 0.8 Protocol specification document at https://cwiki.apache.org/confluence/display/KAFKA/A+Guide+To+The+Kafka+Protocol

The Connection Object

Clients use the Connection object to communicate with the Apache Kafka cluster. The Connection object is an interface layer between your application code and the Apache Kafka cluster.

Connection object is required to create instances of classes Kafka::Producer or Kafka::Consumer.

Kafka Connection API is implemented by Kafka::Connection class.

    use Kafka::Connection;

    # connect to local cluster with the defaults
    my $connection = Kafka::Connection->new( host => 'localhost' );

The main attributes of the Connection object are:

The IO Object

The Kafka::Connection object use internal class Kafka::IO to maintain communication with the particular server of Kafka cluster The IO object is an interface layer between Kafka::Connection object and the network.

Kafka IO API is implemented by Kafka::IO class. Note that end user normally should have no need to use Kafka::IO but work with Kafka::Connection instead.

    use Kafka::IO;

    # connect to local server with the defaults
    my $io = Kafka::IO->new( host => 'localhost' );

The main attributes of the IO object are:

The Producer Object

Kafka producer API is implemented by Kafka::Producer class.

    use Kafka::Producer;

    #-- Producer
    my $producer = Kafka::Producer->new( Connection => $connection );

    # Sending a single message
    $producer->send(
        'mytopic',          # topic
        0,                  # partition
        'Single message'    # message
    );

    # Sending a series of messages
    $producer->send(
        'mytopic',          # topic
        0,                  # partition
        [                   # messages
            'The first message',
            'The second message',
            'The third message',
        ]
    );

The main methods and attributes of the producer request are:

The Consumer Object

Kafka consumer API is implemented by Kafka::Consumer class.

    use Kafka::Consumer;

    $consumer = Kafka::Consumer->new( Connection => $connection );

The request methods of the consumer object are offsets() and fetch().

offsets method returns a reference to the list of offsets of received messages.

fetch method returns a reference to the list of received Kafka::Message objects.

    use Kafka qw(
        $DEFAULT_MAX_BYTES
        $DEFAULT_MAX_NUMBER_OF_OFFSETS
        $RECEIVE_EARLIEST_OFFSETS
    );

    # Get a list of valid offsets up to max_number before the given time
    my $offsets = $consumer->offsets(
        'mytopic',                      # topic
        0,                              # partition
        $RECEIVE_EARLIEST_OFFSETS,      # time
        $DEFAULT_MAX_NUMBER_OF_OFFSETS  # max_number
    );
    say "Received offset: $_" foreach @$offsets;

    # Consuming messages
    my $messages = $consumer->fetch(
        'mytopic',                      # topic
        0,                              # partition
        0,                              # offset
        $DEFAULT_MAX_BYTES              # Maximum size of MESSAGE(s) to receive
    );
    foreach my $message ( @$messages ) {
        if( $message->valid ) {
            say 'payload    : ', $message->payload;
            say 'key        : ', $message->key;
            say 'offset     : ', $message->offset;
            say 'next_offset: ', $message->next_offset;
        } else {
            say 'error      : ', $message->error;
        }
    }

See Kafka::Consumer for additional information and documentation about class methods and arguments.

The Message Object

Kafka message API is implemented by Kafka::Message class.

    if( $message->valid ) {
        say 'payload    : ', $message->payload;
        say 'key        : ', $message->key;
        say 'offset     : ', $message->offset;
        say 'next_offset: ', $message->next_offset;
    } else {
        say 'error      : ', $message->error;
    }

Methods available for Kafka::Message object :

The Exception Object

A designated class Kafka::Exception is used to provide a more detailed and structured information when error is detected.

The following attributes are declared within Kafka::Exception: code, message.

Additional subclasses of Kafka::Exception designed to report errors in respective Kafka classes: Kafka::Exception::Connection, Kafka::Exception::Consumer, Kafka::Exception::IO, Kafka::Exception::Int64, Kafka::Exception::Producer.

Authors suggest using of Try::Tiny's try and catch to handle exceptions while working with Kafka module.

EXPORT ^

None by default.

Additional constants

Additional constants are available for import, which can be used to define some type of parameters, and to identify various error cases.

$KAFKA_SERVER_PORT

default Apache Kafka server port - 9092.

$REQUEST_TIMEOUT

1.5 sec - timeout in secs, for gethostbyname, connect, blocking receive and send calls (could be any integer or floating-point type).

$DEFAULT_MAX_BYTES

1MB - maximum size of message(s) to receive.

$SEND_MAX_RETRIES

4 - The leader may be unavailable transiently, which can fail the sending of a message. This property specifies the number of retries when such failures occur.

$RECEIVE_MAX_RETRIES

4 - The leader may be unavailable transiently, which can fail the receiving of a response. This property specifies the number of retries when such failures occur.

$RETRY_BACKOFF

200 - (ms)

According to Apache Kafka documentation:

Producer Configs - Before each retry, the producer refreshes the metadata of relevant topics. Since leader election takes a bit of time, this property specifies the amount of time that the producer waits before refreshing the metadata.

Consumer Configs - Backoff time to wait before trying to determine the leader of a partition that has just lost its leader.

$RECEIVE_LATEST_OFFSET

-1 : special value that denotes latest available offset.

$RECEIVE_EARLIEST_OFFSETS

-2 : special value that denotes earliest available offset.

$DEFAULT_MAX_NUMBER_OF_OFFSETS

100 - maximum number of offsets to retrieve.

$MIN_BYTES_RESPOND_IMMEDIATELY

The minimum number of bytes of messages that must be available to give a response.

0 - the server will always respond immediately.

$MIN_BYTES_RESPOND_HAS_DATA

The minimum number of bytes of messages that must be available to give a response.

1 - the server will respond as soon as at least one partition has at least 1 byte of data or the specified timeout occurs.

$NOT_SEND_ANY_RESPONSE

Indicates how many acknowledgements the servers should receive before responding to the request.

0 - the server does not send any response.

$WAIT_WRITTEN_TO_LOCAL_LOG

Indicates how long the servers should wait for the data to be written to the local long before responding to the request.

1 - the server will wait the data is written to the local log before sending a response.

$BLOCK_UNTIL_IS_COMMITTED

Wait for message to be committed by all sync replicas.

-1 - the server will block until the message is committed by all in sync replicas before sending a response.

$DEFAULT_MAX_WAIT_TIME

The maximum amount of time (ms) to wait when no sufficient amount of data is available at the time the request is dispatched.

100 - the server will block until the message is committed by all in sync replicas before sending a response.

$MESSAGE_SIZE_OVERHEAD

26 - size of protocol overhead (data added by protocol) for each message.

Compression

According to Apache Kafka documentation:

Kafka currently supports two compression codecs with the following codec numbers:

$COMPRESSION_NONE

None = 0

$COMPRESSION_GZIP

GZIP = 1

$COMPRESSION_SNAPPY

Snappy = 2

Error codes

Possible error codes (complies with a hash of descriptions $ERROR):

$ERROR_MISMATCH_ARGUMENT

-1000 - Invalid argument

$ERROR_CANNOT_SEND

-1001 - Can't send

$ERROR_SEND_NO_ACK

-1002 - No acknowledgement for sent request

ERROR_CANNOT_RECV

-1003 - Can't recv

ERROR_CANNOT_BIND

-1004 - Can't bind

$ERROR_METADATA_ATTRIBUTES

-1005 - Unknown metadata attributes

$ERROR_UNKNOWN_APIKEY

-1006 - Unknown ApiKey

$ERROR_CANNOT_GET_METADATA

-1007 - Can't get Metadata

$ERROR_LEADER_NOT_FOUND

-1008 - Leader not found

$ERROR_MISMATCH_CORRELATIONID

-1009 - Mismatch CorrelationId

$ERROR_NO_KNOWN_BROKERS

-1010 - There are no known brokers

$ERROR_REQUEST_OR_RESPONSE

-1011 - Bad request or response element

$ERROR_TOPIC_DOES_NOT_MATCH

-1012 - Topic does not match the requested

$ERROR_PARTITION_DOES_NOT_MATCH

-1013 - Partition does not match the requested

$ERROR_NOT_BINARY_STRING

-1014 - Not binary string

$ERROR_COMPRESSION

-1014 - Compression error

Contains the descriptions of possible error codes obtained via ERROR_CODE box of Apache Kafka Wire Format protocol response.

$ERROR_NO_ERROR

0 - q{}

No error

$ERROR_UNKNOWN

-1 - An unexpected server error

$ERROR_OFFSET_OUT_OF_RANGE

1 - The requested offset is outside the range of offsets available at the server for the given topic/partition

$ERROR_INVALID_MESSAGE

2 - Message contents does not match its control sum

$ERROR_UNKNOWN_TOPIC_OR_PARTITION

3 - Unknown topic or partition

$ERROR_INVALID_MESSAGE_SIZE

4 - Message has invalid size

$ERROR_LEADER_NOT_AVAILABLE

5 - Unable to write due to ongoing Kafka leader selection

This error is thrown if we are in the middle of a leadership election and there is no current leader for this partition, hence it is unavailable for writes.

$ERROR_NOT_LEADER_FOR_PARTITION

6 - Server is not a leader for partition

Client attempts to send messages to a replica that is not the leader for given partition. It usually indicates that client's metadata is out of date.

$ERROR_REQUEST_TIMED_OUT

7 - Request time-out

Request exceeds the user-specified time limit for the request.

$ERROR_BROKER_NOT_AVAILABLE

8 - Broker is not available

This is not a client facing error and is used only internally by intra-cluster broker communication.

$ERROR_REPLICA_NOT_AVAILABLE

9 - Replica not available

According to Apache Kafka documentation: 'What is the difference between this and LeaderNotAvailable?'

$ERROR_MESSAGE_SIZE_TOO_LARGE

10 - Message is too big

The server has a configurable maximum message size to avoid unbounded memory allocation. This error is thrown when client attempts to produce a message larger than possible maximum size.

$ERROR_STALE_CONTROLLER_EPOCH_CODE

11 - Stale Controller Epoch Code

According to Apache Kafka documentation: '???'

$ERROR_OFFSET_METADATA_TOO_LARGE_CODE

12 - Specified metadata offset is too big

If you specify a value larger than configured maximum for offset metadata.

%ERROR

Contains the descriptions for possible error codes.

BITS64

Know you are working on 64 or 32 bit system

An Example ^

    use 5.010;
    use strict;
    use warnings;

    use Scalar::Util qw(
        blessed
    );
    use Try::Tiny;

    use Kafka qw(
        $KAFKA_SERVER_PORT
        $REQUEST_TIMEOUT
        $RECEIVE_EARLIEST_OFFSETS
        $DEFAULT_MAX_NUMBER_OF_OFFSETS
        $DEFAULT_MAX_BYTES
    );
    use Kafka::Connection;
    use Kafka::Producer;
    use Kafka::Consumer;

    my ( $connection, $producer, $consumer );
    try {

        #-- Connection
        $connection = Kafka::IO->new( host => 'localhost' );

        #-- Producer
        $producer = Kafka::Producer->new( Connection => $connection );

        # Sending a single message
        $producer->send(
            'mytopic',                      # topic
            0,                              # partition
            'Single message'                # message
        );

        # Sending a series of messages
        $producer->send(
            'mytopic',                      # topic
            0,                              # partition
            [                               # messages
                'The first message',
                'The second message',
                'The third message',
            ]
        );

        #-- Consumer
        $consumer = Kafka::Consumer->new( Connection => $connection );

        # Get a list of valid offsets up max_number before the given time
        my $offsets = $consumer->offsets(
            'mytopic',                      # topic
            0,                              # partition
            $RECEIVE_EARLIEST_OFFSETS,      # time
            $DEFAULT_MAX_NUMBER_OF_OFFSETS  # max_number
        );

        if( @$offsets ) {
            say "Received offset: $_" foreach @$offsets;
        } else {
            warn "Error: Offsets are not received\n";
        }

        # Consuming messages
        my $messages = $consumer->fetch(
            'mytopic',                      # topic
            0,                              # partition
            0,                              # offset
            $DEFAULT_MAX_BYTES              # Maximum size of MESSAGE(s) to receive
        );

        if ( $messages ) {
            foreach my $message ( @$messages ) {
                if( $message->valid ) {
                    say 'payload    : ', $message->payload;
                    say 'key        : ', $message->key;
                    say 'offset     : ', $message->offset;
                    say 'next_offset: ', $message->next_offset;
                } else {
                    say 'error      : ', $message->error;
                }
            }
        }

    } catch {
        if ( blessed( $_ ) && $_->isa( 'Kafka::Exception' ) ) {
            warn 'Error: (', $_->code, ') ',  $_->message, "\n";
            exit;
        } else {
            die $_;
        }
    };

    # Closes and cleans up
    undef $consumer;
    undef $producer;
    undef $connection;

DEPENDENCIES ^

In order to install and use this package you will need Perl version 5.10 or later. Some modules within this package depend on other packages that are distributed separately from Perl. We recommend that you have the following packages installed before you install Kafka:

    Const::Fast
    Clone
    Exception::Class
    List::MoreUtils
    Params::Util
    Scalar::Util::Numeric
    String::CRC32
    Sys::SigAction
    Try::Tiny

Kafka package has the following optional dependencies:

    Capture::Tiny
    Config::IniFiles
    Data::Compare
    Proc::Daemon
    Proc::ProcessTable
    Sub::Install
    Test::Deep
    Test::Exception
    Test::TCP

If the optional modules are missing, some "prereq" tests are skipped.

DIAGNOSTICS ^

Debug output can be enabled by setting level via one of the following environment variables:

PERL_KAFKA_DEBUG=1 - debug is enabled for the whole Kafka package.

PERL_KAFKA_DEBUG=IO:1 - enable debug only for Kafka::IO only.

PERL_KAFKA_DEBUG=Connection:1 - enable debug only for particular Kafka::Connection.

It's possible to set different debug levels, like in the following example:

PERL_KAFKA_DEBUG=Connection:1,IO:2

See documentation for a particular module for explanation of various debug levels.

BUGS AND LIMITATIONS ^

Producer and Consumer methods only work with one topic and one partition at a time. Also module does not implement the Offset Commit/Fetch API.

Producer's, Consumer's, Connection's string arguments must be binary strings. Using Unicode strings may cause an error or data corruption.

This module does not support Kafka protocol versions earlier than 0.8.

Kafka::IO->new' uses Sys::SigAction and alarm() to limit some internal operations. This means that if an external alarm() was set, signal delivery may be delayed.

With non-empty timeout, we use alarm() internally in Kafka::IO and try preserving existing alarm() if possible. However, if Time::HiRes::ualarm() is set before calling Kafka modules, its behaviour is unspecified (i.e. it could be reset or preserved etc.).

For gethostbyname operations the non-empty timeout is rounded to the nearest greater positive integer; any timeouts less than 1 second are rounded to 1 second.

You can disable the use of alarm() by setting timeout => undef in the constructor.

The Kafka package was written, tested, and found working on recent Linux distributions.

There are no known bugs in this package.

Please report problems to the "AUTHOR".

Patches are welcome.

MORE DOCUMENTATION ^

All modules contain detailed information on the interfaces they provide.

SEE ALSO ^

The basic operation of the Kafka package modules:

Kafka - constants and messages used by the Kafka package modules.

Kafka::Connection - interface to connect to a Kafka cluster.

Kafka::Producer - interface for producing client.

Kafka::Consumer - interface for consuming client.

Kafka::Message - interface to access Kafka message properties.

Kafka::Int64 - functions to work with 64 bit elements of the protocol on 32 bit systems.

Kafka::Protocol - functions to process messages in the Apache Kafka's Protocol.

Kafka::IO - low-level interface for communication with Kafka server.

Kafka::Exceptions - module designated to handle Kafka exceptions.

Kafka::Internals - internal constants and functions used by several package modules.

A wealth of detail about the Apache Kafka and the Kafka Protocol:

Main page at http://kafka.apache.org/

Kafka Protocol at https://cwiki.apache.org/confluence/display/KAFKA/A+Guide+To+The+Kafka+Protocol

SOURCE CODE ^

Kafka package is hosted on GitHub: https://github.com/TrackingSoft/Kafka

AUTHOR ^

Sergey Gladkov, <sgladkov@trackingsoft.com>

CONTRIBUTORS ^

Alexander Solovey

Jeremy Jordan

Sergiy Zuban

Vlad Marchenko

COPYRIGHT AND LICENSE ^

Copyright (C) 2012-2013 by TrackingSoft LLC.

This package is free software; you can redistribute it and/or modify it under the same terms as Perl itself. See perlartistic at http://dev.perl.org/licenses/artistic.html.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

syntax highlighting: