Status

Current state:  Under Discussion 

Discussion thread: here 

JIRA: KAFKA-4208

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

Motivation

This KIP tries to address the following issues in Kafka.

In most message systems (JMS, QPID etc), streaming systems and most transport systems(HTTP, TCP), it is typical to have a concept of headers and payload.

The payload is traditionally for the business object, and headers are traditionally used for transport routing, filtering etc. Headers are most typically key=value pairs.

In its current state Kafka does not support the ability to have headers natively in its message/record format.

Examples where having separate supported custom headers becomes useful (this is not an exhaustive list).

  • Automated routing of messages based on header information between clusters
  • Enterprise APM tools (e.g. Appdynamics, Dynatrace) need to stitch in 'magic' transaction ids for them to provide end to end transaction flow monitoring.
  • Audit metadata to be recorded with the message, e.g. clientId that produced the record, unique message id, originating clusterId the message was first produced into for multi cluster routing. 
  • Business payload needs to be end to end encrypted and signed without tamper, but eco-system components need access to metadata to achieve tasks.  

Kafka currently has Record<K, V> structure which originally could be used to follow this semantic where by K could contain the headers information, and the V could be the payload.

  • Since message compaction feature it is no longer possible to add metadata to K, else compaction would treat each message as a different keyed message .
  • It is not currently possible to use value part and use some form of a wrapper e.g. Message<H, V>, as for compaction to perform a delete a record is sent with a NULL value, as such for where a delete record is sent using a message wrapper to carry the metadata would not work, as the value technically would no longer be null.

This issue has been flagged by many people over the past period in forums.

Further details and a more detailed case for headers can be seen here : A Case for Kafka Headers

Public Interfaces

This KIP has the following public interface changes:

  1. Add a new headers length and value (byte[]) to the core message format.

  2. Create a Headers class 

    1. Add a headers (MultiMap<String, Object>) field to Headers class

    2. Add accessor methods on the Headers class - void put<T>(String) and a Collection<Object> get(String)

  3. Add a headers field to ProducerRecord and ConsumerRecord. 

  4. Add accessor methods on the Producer/ConsumerRecord Headers getHeaders()
    1. Add ProduceRequest/ProduceResponse V4 which uses the new message format.
  5. Add FetchRequest/FetchResponse V4 which uses the new message format.
  6. The serialisation of the [String, byte[]] header array will on the wire using a strict format
  7. Each headers value will be custom serialisable by the interceptors/plugins that use the header.

For more detail information of the above changes, please refer to the Proposed Changes section.

Proposed Changes

There are four options proposed before this proposal. This details our proposed solution of Option 1 described here. The other options are in the Rejected Alternatives section.

 

The advantages of this proposal are:
  • Adds the ability for producers to set standard header key=value value pairs
  • No incompatible client api change (only new methods)
  • Allows users to specify the serialisation of the header value per header
  • Provides a standardised interface to eco systems of tools that then can grow around the feature

The disadvantage of this proposal is:

  • Change to the message object

Create a Headers Interface and Implementation to encapsulate headers protocol.

This lazy initialises/deserialises, on first method access.

  • Fields:
    • MultiMap<String, Object> headers
  • Constructors
    • ()
    • (bytes[] headerBytes)
  • Methods
    • void putBoolean(String, boolean)
    • void putByte(String, byte)
    • void putChar(String, char)
    • void putShort(String, short)
    • void putInteger(String, integer)
    • void putLong(String, long)
    • void putFloat(String, float)
    • void putDouble(String, double)
    • void putString(String, String)
    • void putBytes(String, byte[])

    • Collection<Object> get(String) – object will be the primitive type wrapper
    • Collection<String> keys()
    • byte[] asBytes()
  1. Add a new Typed class
    1. To hold the typed representations

Add a headers field Headers to both ProducerRecord and ConsumerRecord

  • Accessor methods of Headers getHeaders() added to interface of the ProducerRecord/ConsumerRecord.

Wire protocol change - use attribute bit4 as flag for if headers present. and add (optional) header size and field to the message format

The below is for the core Message wire protocol change needed to fit headers into the message.

A key to this is to ensure the headers cause no overhead if not present.

  • The attributes flag bit is used to keep the message size the same as before if no headers are used

MessageAndOffset => Offset MessageSize Message
  Offset => int64  
  MessageSize => int32
  
  Message => Crc MagicByte Attributes Timestamp KeyLength Key HeadersLength Headers ValueLength Value
    Crc => int32
    MagicByte => int8  <---------------------- Bump up magic byte to 2
    Attributes => int8 <---------------------- Use Bit 4 as boolean flag for if headers present
    Timestamp => int64
    KeyLength => int32
    Key => bytes
    (optional) HeadersLength => int32 <--------------- NEW [optional] size of the byte[] of the serialised headers if headers
    (optional) Headers => bytes <--------------------- NEW [optional] serialised form of the headers Map<int, byte[]>
    ValueLength => int32
    Value => bytes


Wire protocol of the headers bytes (if above mentioned attributes bit flag is true)

The below is for the headers wire protocol.

  • A key can occur multiple times, clients should expect to handle this, this can be represented as a multi map, or an array of values per key in clients.

Headers (bytes) => Array(KeyLength, Key, ValueLength, Value)
  Set =>
	KeyLength => int32 <-----------------NEW size of the byte[] of the serialised key value
	Key => bytes <---------------------- NEW serialised string (UTF8) bytes of the header key
    ValueLength => int32 <-------------- NEW size of the byte[] of the serialised header value
    Value => bytes <-------------------- NEW serialised form of the typed header value -> TypeValue interpretation of this below.
 
TypedValue (bytes) => (Type, Value)
	Type => byte <---------------------- 0x00 = boolean
                                         0x01 = byte
                                         0x02 = char
                                         0x03 = short (int16)
                                         0x04 = int (int32)
                                         0x05 = long (int64)
                                         0x06 = float
                                         0x07 = double
                                         0x08 = string
                                         0x09 = byte[]
    Value => bytes <-------------------- NEW byte array holding the corresponding value in byte array form, 
										 0x00-0x002 - null and boolean this will be zero length

Compatibility, Deprecation, and Migration Plan

  • Current client users should not be affected, this is new api methods being added to client apis
  • Message version allows clients expecting older message version, would not be impacted 
    • older producers simply would produce a message without headers
      • new consumer would simply get a message with empty headers
    • older consumers simply would consume a message oblivious to there being any headers
  • Message version migration would be handled as like in KIP-32

Out of Scope

Some additional features/benefits were noticed and discussed on the above but are deemed out of scope and should be tackled by further KIPS.

  • Core message size reduction
    • remove overhead of 4 bytes for KeyLength when no headers using attributes bit
    • reduce overhead of 4 bytes for KeyLength by using variable length encoded int
    • reduce overhead of 4 bytes for ValueLength by using variable length encoded int
  • Broker side interceptors
    • with headers we could start introducing broker side message interceptors to append meta data or handle messages
  • Single Record consumer API
    • There is many uses cases where single record consumer/listener api is more user friendly - this is evident by the fact spring kafka have already created a wrapper, it would be good to support this natively.

Rejected Alternatives

Map<Int, byte[]> Headers added to the Producer/ConsumerRecord

The concept is similar to the above proposed but int keys

  • Benefits
    • more compact much reduced byte size overhead only 4 bytes.
    • String keys can dwarf the value in byte size.
  • Disadvantages
    • String keys are more common in many systems
    • Requires management of the int key space, where as string keys have natural key space management.
Map<String, String> Headers added to the ConsumerRecord

The concept is similar to the above proposed but with a few more disadvantages.

  • Benefits
    • Adds the ability for producers to set standard header key=value string value pairs
    • No incompatible client api change (only new methods)
    • Allows users to specify the serialisation of the key=value map (String(&=), JSON, AVRO).
    • Provides a standardised interface to eco systems of tools can grow around the feature
  • Disadvantages
    • Change to the message object
    • String key cause a large key, this can cause a message payload thats of 60bytes to be dwarfed by its headers
    • String value again doesn't allow for compact values, and restricts that a value must be a String
    • Not able to use headers on the broker side with custom serialisation
ProducerRecord<K, H, V>, ConsumerRecord<K, H, V>

The proposed change is that headers are Map<int, byte[]> only, this alternative is that headers can be of any type denoted by H

  • Benefits
    • Complete customisation of what a header is.
  • Disadvatages
    • As generics don't allow for default type, this would cause breaking client interface compatibility if done on Producer/ConsumerRecord.
      • Possible work-around would be to have HeadersProducer/ConsumerRecord<K, H, V> that then Producer/ConsumerRecord extend where H is object, this though becomes ugly fast if kept for a time or would require a deprecation / refactor v2 period.
Common Value Message Wrapper - Message<V>

This builds on the status quo and addresses some core issues, but fails to address some more advanced and future use cases and also has some compatibility issues for upgrade/clients not supporting.

please see: Headers Value Message Wrapper

 

Status Quo - Keep Custom Value Message Wrapper - Message<H, P>

This concept is the current defacto way many users are having to temporally deal with the situation, but has some core key issues that it does not resolve.

  • Benefits
    • This will cause no broker side changes to handle the message
  • Disadvantages
    • This would not work with compaction where headers are needed to be sent on delete record which then would not deliver on many of the requirements.
    • Every organization has custom solution
      • No ecosystem of tooling can evolve
      • Cost of Third party vendors integration high, as custom for every organization they integrate for.
    • Not able to make use of headers server side
    • Couple Serialisation and Deserialisation of the value for both the header and payload.

  • No labels