Status

Current state: Under DiscussionAccepted

Discussion thread: here [Change the link from the KIP proposal email archive to your own email thread] 

JIRA: 

Released: 0.11.0.0JIRA: KAFKA-4743 

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

Motivation

This KIP wants to introduce a command-line tool to enable users to reset Consumer Group offsets.

Today, if users want to reprocess old records (i.e. records where offset is smaller than current consumer offset), they will need to modify client-side code (eg. using "KafkaConsumer#seek()" operations).

This process will involve involve write their own program which starts a consumer and commits offset or change client-side code, redeploy the application, and eventually rollback changes in application to consume records as usual. This is a cumbersome process given that users have to be aware of how many partitions are by topic, in which offset they are, and to which offset they want to move to.

...

* Some messages with earlier timestamp might also be consumed."

Public Interfaces

Consumer Group Reset Offset

...

options

Main option

--

...

--verify: This operation will validate that Consumer Group Offsets values defined in the Reset Plan file are valid in the cluster.

...

--execute: This operation will reset Consumer Group Offsets using values defined in the Reset Plan file.

reset-offsets

This option should be executed independently from other Consumer Group options (list, describe, delete, etc.) and it will only support new consumers.

Required Arguments

Scenarios

At least one of the scenarios should be defined to proceed with the execution

Additionally, this tool with support an additional operation to generate and execute in the one step:

• --generate-and-execute

Generate Reset Plan file

To reset Consumer Group offsets first we will need a Reset Plan file with the specific offsets by partitions to reset to. This file could be generated in 3 ways:

...

--reset-to-datetime YYYY-MM-DDTHH:mm:SS.sss

...

...

Default

This process will require 3 arguments:

• --group: defines the Consumer Group ID (i.e. ´group.id´). Required.

• --topics: comma-separated list of topics. Optional, if it’s not specified it will use all topics consumed by Consumer Group.

• --output-file: relative/absolute path to an output file that will be populated with JSON file generated. Optional, if it’s not specified, JSON result will be printed in the terminal.

Reset Plan File: JSON Format

{
	“offsets”: {
		“group_id”: “ConsumerGroup1”,
		“topics”: [
		“topic1”: [
{
“partition”: 0,
		“offset”: 0
	},
		{
“partition”: 1,
		“offset”: 0
	}
],
		“topic2”: [
{
“partition”: 0,
		“offset”: 0
	},
		{
“partition”: 1,
		“offset”: 0
	}
],
		]
},
“version”: 1
}

Reset to Timestamp

This option will support a DateTime format (https://www.w3.org/TR/xmlschema-2/#dateTime) as input. This value will be transformed to epoch milliseconds and then used to get offsets by partition.
 

Pros:

• This options will be the most precise using timestamp, as you move to a specific point in time.

Cons:

• This option will require users to know exactly where they want to move.

Command:

´´´

bin/kafka-consumer-group-reset.sh --generate --group ConsumerGroup1 --topics topic1 --reset-to-timestamp 2017-01-01T09:00:00.000 --output kafka-consumer-group-reset.json

´´´

Reset From Period

This options will support Duration format (https://www.w3.org/TR/xmlschema-2/#duration) as input. This value will be subtracted from current timestamp from client’s machine to obtain the epoch milliseconds to be used to get offsets by partition.

Pros:

• This option will be the easiest to use, as it does not require to know a specific timestamp, but give a point in time depending on when you execute the command.

• Useful for debugging/development purposes where you just need to reprocess a bunch of records from an previous period.

Cons:

• This option will be less accurate than previous one.

Command:

´´´

bin/kafka-consumer-group-reset.sh --generate --group ConsumerGroup1 --topics topic1 --reset-from-period P1M --output kafka-consumer-group-reset.json

´´´

Default

This option will generate a file with the current Consumer Groups offsets by partition. The result of this execution will generate a JSON file that could be used as a template to modify the offsets manually for all partitions.

This option will be useful for clusters prior to release 0.10.0.0, to create an initial file where they can specify specific offsets later. And also could be use as a fixed point in time before starting the application and later reset to this point.

Pros:

• Easiest way to generate a JSON file with the right format.

Cons:

• Could be cumbersome to users to modify specific offsets by partition.

Command:

´´´

bin/kafka-consumer-group-reset.sh --generate  --group ConsumerGroup1 --topics topic1 --output kafka-consumer-group-reset.json

´´´

Verify JSON file

Once JSON file is generated, the tool will offer an option to validate it with the current state of the cluster. The tool will validate the following:

• Topics exist

• Partitions exist

• Offsets exist

• Latest/Current/Oldest offset

If topics or partitions don’t exist in the cluster, tool will advise that those topics/partitions won’t be processed.

If offset is not available anymore (e.g. retention remove it from the log), tool will advise that will reset the offset to the oldest or latest (i.e. resetting the offset using ‘auto.offset.reset’ policy).

Latest/Current/Oldest offset information will give the user the idea about where in the log the offset will be set.

This process will require 1 argument:

• --reset-json-file: relative or absolute path to JSON file generated/customized.

Command:

´´´

bin/kafka-consumer-group-reset.sh --verify --reset-json-file kafka-consumer-group-reset.json

´´´

Execute JSON file

Once user feels comfortable with the Reset Plan file, could prepare the environment to execute the file.

To execute the reset process, the tool will require and validate that the Consumer Group become inactive first, to avoid inconsistencies in the client side.

This process will require 1 argument:

• --reset-json-file: relative or absolute path to JSON file generated/customized.

Command:

´´´

bin/kafka-consumer-group-reset.sh --execute --reset-json-file kafka-consumer-group-reset.json

´´´

Scopes

The scopes will be defined extending the existing `topic` argument, using the following format:

--topic <topic-name>:<partition numbers>
ex: --topic topic1 --topic topic2:0,1,2

Execution Options

TOPIC                 PARTITION NEW-OFFSET NEW-LAG LOG-END-OFFSET CONSUMER-ID HOST CLIENT-ID
foo                   0         90         10      100            -           -    -

Reset Plan File: CSV Format

The CSV schema will consist on:

<topic>,<partition>,<offset>

Sample:

topic-1,0,100
topic-1,1,200
topic-1,2,0
topic-2,0,0

Implementation Details 

These options will be implemented inside the `ConsumerGroupCommand`. The `resetOffsets` operation will looks like this:

def resetOffsets(): Map[PartitionAssignmentState, Long] = {
  val groupId = opts.options.valueOf(opts.groupOpt)
  val (state, assignments) = describeGroup() //(1)
  assignments match {
    case None =>
      // applies to both old and new consumer
      printError(s"The consumer group '$groupId' does not exist.")
      Map.empty
    case Some(assignments) =>
      state match {
        case Some("Dead") =>
          printError(s"Consumer group '$groupId' does not exist.")
          Map.empty
        case Some("Empty") => //(2)
          val assignmentsToReset = getAssignmentsToReset(assignments) //(3)
          val assignmentsPrepared = prepareAssignmentsToReset(assignmentsToReset) //(4)
          val execute = opts.options.has(opts.executeOpt)
          if(execute) //(5)
            resetAssignments(assignmentsPrepared)
          assignmentsPrepared
        case Some("PreparingRebalance") | Some("AwaitingSync") =>
          printError(s"Consumer group '$groupId' offsets cannot be reset if it is rebalancing.")
          Map.empty
        case Some("Stable") =>
          printError(s"Consumer group '$groupId' offsets cannot be reset if it has members active.")
          Map.empty
        case other =>
          // the control should never reach here
          throw new KafkaException(s"Expected a valid consumer group state, but found '${other.getOrElse("NONE")}'.")
      }
  }
}

(1) It will get assignments from `describeGroup` operation

(2) Then use `getAssignmentsToReset` to filter the assignments with the values defined by `–topic` of or `--all-topics`.

(3) Then get new offset by assignments using `prepareAssignmentsToReset`.

(4) It will be only executed when the ConsumerGroup selected is inactive o avoid race conditions. 

(5) It will be executed only if it is asked explictly. This will consist in create a Consumer using the same `group.id` and use:

consumer.assign(List(topicPartition).asJava)
consumer.seek(topicPartition, offset)
consumer.commitSync()

To change the offsetsFinally user can use the ´verify´ option to validate that current and defined offset are the same.

Compatibility, Deprecation, and Migration Plan

This KIP won’t require a Migration Plan.

‘reset-to-datetime’ and ‘reset-fromto-period’ will require Timestamp Index, which will make them only available from version 0.10.1.0. 

Nevertheless, it is possible to achieve the same result since release 0.10.0 using timestamp metadata by record, doing a linear search starting from latest. However, we are not considering to support this option because it will be an expensive/slow option.

All the other options will be available for releases from 0.10.0.0 given that KIP-97 is implemented.

...

This KIP will not support old consumers that store offsets in Zookeeper.

Test Plan

...

• A unit test to validate

...

• that --reset-offset is executed independently from other Consumer Group options (list, delete, describe, etc.)
• A unit test to validate

...

Generate:

...

• that only one scenario is specified

...

Validate that current offsets by partition are obtained when default option is executed

...

Validate that an offset is obtained when a datetime is specified

...

Validate that oldest offset is obtained when a datetime is older than the oldest timestamp by partition

...

Validate that latest offset is obtained when a datetime is higher than the latest timestamp by partition

...

Validate that an offset is obtained when a period is specified

...

Validate that oldest offset is obtained when a period minus current timestamp is older than the oldest timestamp by partition

...

Validate that latest offset is obtained when a period minus current timestamp is higher than the latest timestamp by partition

• A unit test to validate that only one scope is specified
• A unit test to validate that only one execution option is specified
• A unit test by combination of scenario, scope and execution option.
• A unit test to validate that when calculated offset by partition is older that the earliest offset, tool resets offset to earliest (e.g. when we specified --reset-to 0, when earliest offset is 10)
• A unit test to validate that when calculated offset by partition is bigger that the latest offset, tool resets offset to latest (e.g. when we specified --reset-to 0, when earliest offset is 10)

Rejected Alternatives

None

(no scenario arguments)

...

Verify:

• Validate that JSON structure is correct

• Validate that topics exist

• Validate that partitions exist

• Validate that topics are actually consumed by Consumer Group

• Warn that topic does not exist and nothing will be executed

• Warn that topic is not been consumed by Consumer Group and nothing will be executed

• Warn that partition does not exist and nothing will be executed.

• Warn that specified offset is out of range (ie. from oldest to latest) and it will be reset to oldest or latest depending on which one is closer.

...

Execute:

• Validate that ConsumerGroup is inactive.

• Validate that offsets are updated in cluster after execution.

• Warn that non-existing/not-consumed topic has not been updated

• Warn that non-existing partitions has not been updated

• Warn that offsets out of bound has been updated to oldest or latest, depending which one is closer.

Rejected Alternatives

None

1. Reset to specific offset: Move Consumer Group to specific offset by partition.

2. Reset to datetime: Find last offset by datetime, and reset offset by partition.

3. Reset by period: Find last offset from datetime calculated from an input period and current timestamp (e.g. 5 days ago.).

This process will be separated in 3 steps:

Generate JSON file with specific offsets by partition to reset

Verify JSON file consistency and current state of consumer group offsets

Execute the JSON file, resetting offsets to values specified.

Also, it will consider an option to generate and execute in one step.

This options will enable clients to move back and forward, depending on where they are currently in the topic.

For instance: if a topic T1 with 1 partition and 100 records (offsets 0-99), has one Consumer Group C1 inactive, with current offset 20. If we reset to a point in time where offset is 80, client will be moving forward. Another Consumer Group C2 that has been actively consuming records from T1, if it’s reset to the same point in time, it will be moving backward.