Status

Current state: DraftUnder Review

Discussion thread: TBD

JIRA: TBD

...

And the Raft node will respond with either an approval or reject to the pre-vote request in the same way as responding to the normal vote request. Each candidate node will first attempt to send pre-votes to the known quorum members. When it gets the majority of pre-votes approved, the candidate will proceed to the actual vote. If pre-vote fails, the active candidate will not continue to attempt to do the actual vote.

...

{
  "apiKey": N,
  "type": "request",
  "name": "ReadOffsetRequest",
  "validVersions": "0",
  "flexibleVersions": "0+",
  "fields": [
    { "name": "ClusterId", "type": "string", "versions": "0+",
      "nullableVersions": "0+", "default": "null"},
    { "name": "ReadTimestamp", "type": "int64", "versions": "1", "default": "-10",
      "about": "The timestamp of the read query."},
    { "name": "Topics", "type": "[]TopicData",
      "versions": "0+", "fields": [
      { "name": "TopicName", "type": "string", "versions": "0+", "entityType": "topicName",
        "about": "The topic name." },
      { "name": "Partitions", "type": "[]PartitionData",
        "versions": "0+", "fields": [
        { "name": "PartitionIndex", "type": "int32", "versions": "0+",
          "about": "The partition index." },
      ]
      }
    ]
    }
  ]
}

...

{
  "apiKey": N,
  "type": "response",
  "name": "ReadOffsetResponse",
  "validVersions": "0",
  "flexibleVersions": "0+",
  "fields": [
    { "name": "ClusterId", "type": "string", "versions": "0+",
      "nullableVersions": "0+", "default": "null"},
    { "name": "Topics", "type": "[]TopicData",
      "versions": "0+", "fields": [
      { "name": "TopicName", "type": "string", "versions": "0+", "entityType": "topicName",
        "about": "The topic name." },
      { "name": "Partitions", "type": "[]PartitionData",
        "versions": "0+", "fields": [
        { "name": "PartitionIndex", "type": "int32", "versions": "0+",
          "about": "The partition index." },
		{ "name": "AppliedOffset", "type": "int64", "versions": "0+",
          "about": "The last applied offset on the leader." }, 
      ]
      }
    ]
    }
  ]
}

Compatibility, Deprecation, and Migration Plan

• What impact (if any) will there be on existing users?
• If we are changing behavior how will we phase out the older behavior?
• If we need special migration tools, describe them here.
• When will we remove the existing behavior?

Rejected Alternatives

Once getting the offset read result, follower/observers would buffer those pending read requests until its state machine has already reached the marked applied offset level, to ensure the read consistency on the follower side. If the query was on hold for too long, we rely on client side to timeout and retry another candidate based off its own timeout, instead of implementing a complex lag tracking mechanism for now.

Client side could optionally implement a flag to decide whether to perform stale read or linearizable read to avoid long holding for the result, which could be done in the follow-up work.

Compatibility, Deprecation, and Migration Plan

This is a new feature which does not affect existing component's behavior. Older client should also be compatible.

Rejected Alternatives

N/AIf there are alternative ways of accomplishing the same thing, what were they? The purpose of this section is to motivate why the design is the way it is and not some other way.

Citation

Please stop calling databases CP or AP, Published by Martin Kleppmann on 11 May 2015