...

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

Motivation

This KIP is trying to resolve the following important issues in Kafka:

...

Today the broker needs to decompress compressed messages, assign offsets to each message and recompress the messages again. This causes additional CPU cost.

...

This KIP is trying to avoid server side recompression.

This KIP is a distilled/improved version of an earlier discussion that we started.

...

MessageAndOffset => MessageSize Offset Message
  MessageSize => int32
  Offset => int64 <------ CHANGE: Base offset for wrapper message of compressed message, relative offset for inner compressed message.
  
  Message => Crc MagicByte Attributes Timestamp KeyLength Key ValueLength Value
    Crc => int32
    MagicByte => int8
    Attributes => int8
    Timestamp => int64 <---------------------- NEW
    KeyLength => int32
    Key => bytes
    ValueLength => int32
    Value => bytes

...

...

Proposed Changes

Wire protocol change

Add a timestamp field to message

• The timestamp will be assigned by broker upon receiving the message. If the message is coming from mirror maker, the original time stamp will be discarded. In essence, we treat timestamp as another offset-like field. If application needs a timestamp, it can be put into payload.
• The timestamp will be used to build the log index.
• The tiimestamp accuracy would be millisecond
• The timestamp of the outer message of compressed messages will be the latest timestamp of all its inner messages.
  • If the compressed message is not compacted, the relative offsets of inner messages will be contiguous and share the same timestamp.
  • If the compressed message is compacted, the relative offsets of inner messages may not be contiguous. Its timestamp will be the timestamp of the last inner message.
• The followers will not reassign timestamp but simply update a in memory lastAppendedTimestamp and append the message to the log.
• To handle leader migration where new leader has slower clock than old leader, all the leader should append max(lastAppendedTimestamp, currentTimeMillis) as the timestamp.

Change the usage of offset field

• When the producer compresses a message, write the relative offset value in the raw message's offset field. Leave the wrapped message's offset blank.
• When broker receives a compressed message, it only needs to
  
  1. Decompress the message to verify the CRC and relative offset.
     NOTE: If the relative offset is not contiguous (e.g., if this is a mirrored compacted topic), the broker will reassign the relative offsets. There are two ways to handle this - (i) reject the ProducerRequest or (ii) just assign the relative offsets. We chose to reassign offsets rather than reject the request because there is a useful use case where mirror maker can do a direct copy from source cluster to destination cluster without even decompressing the message. In this case, the compressed message can have noncontinuous relative offsets (for compacted topics).
  2. Set outer message's base offset. (Since the broker only needs to update the message-set header, there is no need to re-compress message sets.)
• When the log cleaner compacts log segments, it needs to update the inner message's relative offset values. (This will leave "holes" inside the new wrapped message).

Change time based log rolling and retention to use timestamp

Time based log rolling and retention currently use the file creation time and last modified time. This does not work for newly created replicas because the time attributes of files is different from the true message append time. The following changes will address this issue.

• The time based log rolling will be based on the timestamp of the first message in the log segment file.
• The time based log retention will be based on the timestamp of the last message in the log segment file.

New time-based log index

In order to enable timestamp based search at finer granularity, we need to add the timestamp to log indices as well.

Because all the index files are memory mapped files the main consideration here is to avoid significantly increasing the memory consumption.

The time index file needs to be built just like the log index file based on each log segment file.

Use a time index for each log segment to save the timestamp -> log offset at minute granularity

Create another index file for each log segment with name SegmentBaseOffset.time.index to have index at minute level. The time index entry format is:

Time Index Entry => Timestamp Offset
  Timestamp => int64
  Offset => int32

Users don't typically need to look up offsets with seconds granularity.

Compatibility, Deprecation, and Migration Plan

Compatibility, Deprecation, and Migration Plan

NOTE: This part is drafted based on the assumption that KIP-31 and KIP-32 will be implemented in one patch.

The proposed protocol is not backward compatible. The migration plan are as below:

...

For upgraded consumers, they can handle both V0 and V1.

Rejected Alternatives

Add a timestamp field to log index entry

The most straight forward approach to have a time index is to let the log index files have a timestamp associate with each entry.

Log Index Entry => Offset Position Timestamp
  Offset => int32
  Position => int32
  Timestamp => int64

None

 Because the index entry size become 16 bytes instead of 8 bytes. The index file size also needs to be doubled. As an example, one of the broker we have has ~3500 partitions. The index file took about 16GB memory. With this new format, the memory consumption would be 32GB.