Kafka 压缩

概括

需要理解 kafka 压缩则需要理解 Kafka 的存储格式.

Kafka 存储格式

RecordBatch

baseOffset: int64
batchLength: int32
partitionLeaderEpoch: int32
magic: int8 (current magic value is 2)
crc: int32
attributes: int16
    bit 0~2:
        0: no compression
        1: gzip
        2: snappy
        3: lz4
        4: zstd
    bit 3: timestampType
    bit 4: isTransactional (0 means not transactional)
    bit 5: isControlBatch (0 means not a control batch)
    bit 6~15: unused
lastOffsetDelta: int32
firstTimestamp: int64
maxTimestamp: int64
producerId: int64
producerEpoch: int16
baseSequence: int32
records: [Record]

Record

length: varint
attributes: int8
    bit 0~7: unused
timestampDelta: varint
offsetDelta: varint
keyLength: varint
key: byte[]
valueLen: varint
value: byte[]
Headers => [Header]

Record Header

headerKeyLength: varint
headerKey: String
headerValueLength: varint
Value: byte[]

Note: 图片来源. 推荐阅读该文章来更好的了解 Kafka 消息格式演变过程.

消息比较

0000 0000 0000 0000 0000 0040 0000 0000
02e3 0171 9400 0000 0000 0000 0001 6ad9
0153 7e00 0001 6ad9 0153 7eff ffff ffff
ffff ffff ffff ffff ff00 0000 011c 0000
0006 6b65 790a 7661 6c75 6500

0000 0000 0000 0001    0000 0054 0000 0000
02e5 cb48 0600 0100 0000 0000 0001 6ad9
5427 af00 0001 6ad9 5427 afff ffff ffff
ffff ffff ffff ffff ff00 0000 011f 8b08
0000 0000 0000 0093 6160 6060 cb4e ade4
2a4b cc29 4d65 0000 55dc 0454 0f00 0000 

用可视化的方式分析

76B =======================Header============================
0000 0000 0000 0000      =>    first offset              =>    0
0000 0040                =>    length                    =>    64
0000 0000                =>    partition leader epoch    =>    0
02                       =>     magic                    =>    2
e3 0171 94               =>    crc32                     =>    3808522644
00 00                    =>    attributes                =>    0
00 0000 00               =>    last offset delta         =>    0
00 0001 6ad9 0153 7e     =>    first timestamp           =>    1558418903934
00 0001 6ad9 0153 7e     =>    max timestamp             =>    1558418903934
ff ffff ffff ffff ff     =>    producer id               =>    -1
ff ff                    =>    producer epoch            =>    -1
ff ffff ff               =>    first sequence            =>    -1
00 0000 01               =>    record count              =>    1

=======================Records===========================================
1c                        =>    length(变长)                =>    14
00                        =>     arrtibutes                =>    弃用
00                        =>    timestamp delta(变长)       =>    0
00                        =>    offset delta(变长)          =>    0
06                        =>    key length(变长)            =>    3
6b65 79                   =>    key                        =>    "key"
0a                        =>    value length(变长)          =>    5
7661 6c75 65              =>    value                      =>    "value"
00                        =>    headers    counts(变长)     =>    0

开启消息压缩

0000 0000 0000 0000 0000 0040 0000 0000
02e3 0171 9400 0000 0000 0000 0001 6ad9
0153 7e00 0001 6ad9 0153 7eff ffff ffff
ffff ffff ffff ffff ff00 0000 011c 0000
0006 6b65 790a 7661 6c75 6500
================ 上面是 76B 未压缩的数据 =====================
0000 0000 0000 0001        first offset
0000 0054                length
0000 0000                partition leader epoch
02                        magic
e5 cb48 06                crc32
00 01                    attributes
00 0000 00                last offset delta
00 0001 6ad9 5427 af     first timestamp
00 0001 6ad9 5427 af     max timestamp
ff ffff ffff ffff ff     producer id
ff ff                     producer epoch
ff ffff ff                 first sequence
00 0000 01                 record count

1f 8b08 0000 0000 0000 0093 6160 6060 cb4e ade4 2a4b cc29 4d65 0000 55dc 0454 0f00 0000 

结论

消息压缩只是针对 records 部分.