Page History

Authors Satish Duggana, Sriharsha Chintalapani, Satish DugganaYing Zheng, Suresh Srinivas, Ying Zheng (alphabetical order by the last names)

Status

Current State: Discussion "Accepted"

Discussion Thread: Discuss Thread here

JIRA: 

Google doc version of this wiki is located here.

Motivation

Kafka is an important part of data infrastructure and is seeing significant adoption and growth. As the Kafka cluster size grows and more data is stored in Kafka for a longer duration, several issues related to scalability, efficiency, and operations become important to address.

...

The total storage required on a cluster is proportional to the number of topics/partitions, the rate of messages, and most importantly the retention period. A Kafka broker typically has a large number of disks with the a total storage capacity of 10s of TBs. The amount of data locally stored on a Kafka broker presents many operational challenges.

...

Kafka cluster storage is typically scaled by adding more broker nodes to the cluster. But this also adds needless memory and CPUs to the cluster making overall storage cost less efficient compared to storing the older data in external storage. Larger A larger cluster with more nodes also adds to the complexity of deployment and increases the operational costs.

...

In the tiered storage approach, Kafka cluster is configured with two tiers of storage - local and remote. Local The local tier is the same as the current Kafka that uses the local disks on the Kafka brokers to store the log segments. The new remote tier uses systems, such as HDFS or S3 to store the completed log segments. Two separate retention periods are defined corresponding to each of the tiers. With remote tier enabled, the retention period for the local tier can be significantly reduced from days to few hours. The retention period for remote tier can be much longer, days, or even months. When a log segment is rolled on the local tier, it is copied to the remote tier along with the corresponding offset indexindexes. Latency sensitive applications perform tail reads and are served from local tier leveraging the existing Kafka mechanism of efficiently using page cache to serve the data. Backfill and other applications recovering from a failure that needs data older than what is in the local tier are served from the remote tier.

...

It does not support compact topics with tiered storage. Topic created with the effective value for remote.logstorage.enabled as enable as true, can not change its retention policy from delete to compact.

...

• receives callback events for leadership changes and stop/delete events of topic partitions on a broker.
• delegates copy, read, and delete of topic partition segments to a pluggable storage manager(viz RemoteStorageManager) implementation and maintains respective remote log segment metadata through RemoteLogMetadataManager.

`RemoteLogManager` is an internal component and it is not a public API. 

`RemoteStorageManager` is an interface to provide the lifecycle of remote log segments and indexes. More details about how we arrived at this interface are discussed in the document. We will provide a simple implementation of RSM to get a better understanding of the APIs. HDFS and S3 implementation are planned to be hosted in external repos and these will not be part of Apache Kafka repo. This is inline with the approach taken for Kafka connectors.

...

RLM creates tasks for each leader or follower topic partition, which are explained in detail here.

• RLM Leader Task
  • It checks for rolled over LogSegments (which have the last message offset less than last stable offset of that topic partition) and copies them along with their offset/time/transaction/producer-snapshot indexes and leader epoch cache to the remote tier. It also serves the fetch requests for older data from the remote tier. Local logs are not cleaned up till those segments are copied successfully to remote even though their retention time/size is reached.

• RLM Follower Task 
  • It keeps track of the segments and index files on the remote tier by looking into RemoteLogMetdataManager. RLM follower can also serve reading old data from the remote tier.

RLM maintains a bounded cache(possibly LRU) of the index files of remote log segments to avoid multiple index fetches from the remote storage. They are stored in a directory `remote-log-index-cache` under log dir. These indexes can be used in the same way as local segment indexes are used. User can configure `remote.log.index.file.cache.total.size.mb` to set the total size that can be used for these index files. 

The earlier approach consists of pulling the remote log segment metadata from remote log storage APIs as mentioned in the earlier RemoteStorageManager_Old section. This approach worked fine for storages like HDFS. One of the problems of relying on the remote storage to maintain metadata is that tiered-storage needs to be strongly consistent, with an impact not only on the metadata itself (e.g. LIST in S3) but also on the segment data (e.g. GET after a DELETE in S3). Also, the cost (and to a lesser extent performance) of maintaining metadata in remote storage needs to be factored in. In the case of S3, frequent LIST APIs incur huge costs. 

...

The below diagram gives a brief overview of the interaction between leader, follower, and remote log and metadata storagesstorage. It will be described more in detail in the next section.

...

Currently, followers build the auxiliary state (i.e. leader epoch sequence, producer snapshot state) when they fetch the messages from the leader by reading the message batches. Incase of tiered storage, follower finds the offset and leader epoch up to which the auxiliary state needs to be built from the leader. After which, followers start fetching the data from the leader starting from that offset. That offset can be local-log-start-offset or last-tiered-offset. Local-log-start-offset is the log start offset of the local storage. Last-tiered-offset  offset is the offset up to which the segments are copied to remote storage. We will describe pros and cons of choosing these segments.

last-tiered-offset

• The advantage with of this option is that followers can catch up quickly with the leader as the segments that are required to be fetched by followers are the segments that are not yet moved to remote storage.  
• One disadvantage with this approach is that followers may have a few local segments than the leader. When that follower becomes a leader then the existing followers will truncate their logs to the leader's local log-start-offset. 

...

• This will honour local log retention in case of leader switches.
• It will take longer for a lagging follower to become an insync replica by catching up with the leader. One of those cases can be a new follower replica added for a partition need to start fetching from local log start offset to become an insync follower. So, this may take longer based on the local log segments available on the leader. 

We prefer to go with the local log start offset as the offset from which follower starts to replicate the local log segments for the reasons mentioned above mentioned reasons.

With tiered storage, the leader only returns the data that is still in the leader's local storage. Log segments that exist only on remote storage are not replicated to followers as those are already present in remote storage. Followers fetch offsets and truncate their local logs if needed with the current mechanism based on the leader's local-log-start-offset. This is described with several cases in detail in the next section.

...

1) Retrieve the Earliest Local Offset (ELO) and the corresponding leader epoch (ELO-LE) from the leader with a ListOffset request (timestamp = -34)

2) Truncate local log and local auxiliary state

...

2) Fetch the leader epoch sequence sequence  and  producer snapshot from remote storage (using remote storage fetcher thread pool)

...

After building the local leader epoch cache, the follower transfers back to Fetching state, and continues fetching from ELO. We preferred to go with the latter option as it can get the required state from remote storage. 

Let us discuss a few cases that followers can encounter while it tries to replicate from the leader and build the auxiliary state from remote storage.

OMRS OMTS : OffsetMovedToRemoteStorageOffsetMovedToTieredStorage

ELO : Earliest-Local-Offset

...

...

In this case, local segments might have already been deleted because of the local retention settings, or the follower has been offline for a very long time. The follower receives OFFSET_MOVED_TO_TIERED_STORAGE error while trying to fetch the desired offset. The follower has to truncate all the local log segments , because we know the data already expired on the leader.

...

...

...

...

In this case, it is acceptable to lose data, but we have to keep the same behaviour as described in the KIP-101.

After restart, B losses message 1 and 2. B becomes the new leader, and receives a new message 3 (LE1LE-1, offset 1).

(Note: This may not be technically an unclean-leader-election, because B may have not been removed from ISR because both of the 2 brokers crashed at the same time.)

...

A new message (message 4) is received. The 2nd segment on broker B (seg-1-1) is shipped to remote storage.

The Consider the local segments upto up to offset 2 are deleted on both brokers.:

A consumer fetches offset 0, LE-0. According to the local leader epoch cache, offset 0 LE-0 is valid. So, the broker returns message 0 from remote segment 0-1.

A pre-KIP-320 consumer fetches offset 1, without leader epoch info. According to the local leader epoch cache, offset 1 belongs to LE-1. So, the broker returns message 3 from remote segment 1-1, rather than the LE-0 offset 1 message ( message 1 ) in seg-0-1.

A consumer fetches offset 2 LE0 LE-0 is fenced (KIP-320).

A consumer fetches offset 1 LE-1 LE1 receives message 3 from remote segment 1-1.

...

Scenario 5: log divergence in remote storage - unclean leader election

step 1

step 2

An out-of-sync broker B becomes the new leader, after broker A is down. (unclean leader election)

...

...

...

A follower can be considered as a leader by the controller based on its replica configuration. When a follower becomes a leader it needs to find out the offset from which the segments to be copied to remote storage. This is found by traversing from the latest leader epoch from leader epoch history and find the highest offset of a segment with that epoch copied into remote storage. If it can not find an entry then it checks for the previous leader epoch till it finds an entry, If there are no entries till the earliest leader epoch in leader epoch cache then it starts copying the segments from the earliest epoch entry’s offset.

Step 1:

...

RemoteLogManager copies transaction index and producer-id-snapshot along with the respective log segment earlier to last-stable-offset. This is used by the followers to return aborted transactions in fetch requests with isolation level as READ_COMMITTED. // Add how the follower may need to go to offset up to which the producer snapshot exists and start fetching from there. 

Consumer Fetch Requests

For any fetch requests, ReplicaManager will proceed with making a call to readFromLocalLog, if this method returns OffsetOutOfRange exception it will delegate the read call to RemoteLogManager. More details are explained in the RLM/RSM tasks section.

Fetch from follower

There are no changes required for this to work in the case of tiered storage. If the remote storage is not available then it will throw a new error TIERED_STORAGE_NOT_AVAILABLE.

Other APIs

DeleteRecords

There is no change in the semantics of this API. It deletes records until the given offset if possible. This is equivalent to updating logStartOffset of the partition log with the given offset if it is greater than the current log-start-offset and it is less than or equal to high-watermark. If needed, it will clean remote logs asynchronously after updating the log-start-offset of the log. RLMTask for leader partition periodically checks whether there are remote log segments earlier to logStartOffset and the respective remote log segment metadata and data are deleted by using RLMM and RSM.

...

This API is enhanced with supporting new target timestamp value as -3 which 4 which is called EARLIEST_LOCAL_TIMESTAMP. There will not be any new fields added in request and response schemes but there will be a version bump to indicate the version update. This request is about the offset that the followers should start fetching to replicate the local logs. It represents the log-start-offset available in the local log storage which is also called as local-log-start-offset. All the records earlier to this offset can be considered as copied to the remote storage. This is used by follower replicas to avoid fetching records that are already copied to remote tier storage.

When a follower replica needs to fetch the earliest messages that are to be replicated then it sends a request with the target timestamp as EARLIEST_LOCAL_TIMESTAMP. 

...

This is received by RLM to register for new leaders so that the data can be copied to the remote storage. RLMM will also register the respective metadata partitions for the leader/follower partitions if they are not yet subscribed.  

Stopreplica

RLM receives a callback and unassigns the partition for leader/follower task, If the delete option is enabled then the leader will stop RLM task and stop processing and it sets all the remote log segment metadata of that partition with a delete marker and publishes them to RLMM. The controller will not allow topic with the same name to be created till all the segments are cleaned up from remote storage.

It was discussed in the community earlier for adding UUID to represent a topic along with the name as part of KIP-516. This enhancement will be useful to make the deletion of topic partitions in remote storage asynchronously without blocking the creation of topic with the same name even though all the segments are not deleted in remote storage.   

OffsetForLeaderEpoch

Look into leader epoch checkpoint cache. This is stored in tiered storage and it may be fetched by followers from tiered storage as part of the fetch protocol. 

...

After a topic-partition is successfully processed by the thread pool, it's scheduled processing time is set to ( now() + rlm_process_interval_ms ). rlm_process_interval_ remote.log.manager.task.interval.ms ). remote.log.manager.task.interval.ms can be configured in broker config file.

If the process of a topic-partition is failed due to remote storage error, its scheduled processing time is set to ( now() + rlm_retry_interval_ms ). rlm_retry_interval_ms can be configured in broker config fileit follows retry backing off algorithm with intiial retry interval as `remote.log.manager.task.retry.interval.ms`, max backoff as `remote.log.manager.task.retry.backoff.max.ms`, and jitter as `remote.log.manager.task.retry.jitter`. You can see more details about the exponential backoff algorithm here.

When a topic-partition is unassigned from the broker, the topic-partition is not currently processed by the thread pool, the topic-partition is directly removed from the list; otherwise, the topic-partition is marked as "deleted", and will be removed after the current process is done.

...

Handle expired remote segments (leader and follower)

RLM leader computes the log segments to be deleted based on the remote retention config. It updates the earliest offset for the given topic partition in RLMM. It gets all the remote log segment ids and removes them from remote storage using RemoteStorageManager. It also removes respective metadata using RemoteLogMetadataManager.RLM follower fetches the earliest offset for the earliest leader epoch by calling RLMM.earliestLogOffset(TopicPartition topicPartition, int leaderEpoch) and updates that as the log start offset.

2. Remote Storage Fetcher Thread Pool

...

1. find out the corresponding RemoteLogSegmentId from RLMM and startPosition and endPosition from the offset index.
2. try to build Records instance data fetched from RSM.fetchLogSegmentData(RemoteLogSegmentMetadata remoteLogSegmentMetadata, Long startPosition, Long endPosition)
   1. if success, RemoteFetchPurgatory will be notified to return the data to the client
   2. if the remote segment file is already deleted, RemoteFetchPurgatory will be notified to return an error to the client.
   3. if the remote storage operation failed (remote storage is temporarily unavailable), the operation will be retried with Exponential Back-Off, until the original consumer fetch request timeout.

Remote Log

...

Metadata State transitions

Image Removed

Image Added

COPY_SEGMENT_STARTED - This state indicates that the segment copying to remote storage is started but not yet finished.

COPY_SEGMENT_FINISHED - This state indicates that the segment copying to remote storage is finished.

The leader broker copies the log segments to the remote storage and puts the remote log segment metadata with the state as “COPY_SEGMENT_STARTED” Leader broker copies the log segments to the remote storage and puts the remote log segment metadata with the state as “COPY_STARTED” and updates the state as “COPY_FINISHED” “COPY_SEGMENT_FINISHED” once the copy is successful. Leaders also remove the remote log segments based on the retention policy. Before the log segment is removed using RSM.deleteLogSegment(RemoteLogSegmentMetadata remoteLogSegmentMetadata), it updates the remote log segment with the state as DELETE_STARTED and it updates with DELETE_FINISHED once it is successful.

DELETE_SEGMENT_STARTED - This state indicates that the segment deletion is started but not yet finished.

DELETE_SEGMENT_FINISHED - This state indicates that the segment is deleted successfully.

 Leader partitions publish both the above delete segment events when remote log retention is reached for the respective segments. Remote Partition Removers also publish these events when a segment is deleted. 

DELETE_PARTITION_MARKED - This is published when a topic/partition is deleted by the controller. This partition is marked for delete by the controller. That means, all its remote log segments are eligible for deletion so that remote partition removers can start deleting them.

DELETE_PARTITION_STARTED - This state indicates that the partition deletion is started but not yet finished.

DELETE_PARTITION_FINISHED - This state indicates that the partition is deleted successfully.

Remote Partition Removers also publish these events when a partition is deleted. 

When a partition is deleted, the controller updates its state in RLMM with DELETE_PARTITION_MARKED and it expects RLMM will have a mechanism to clean up the remote log segments. This process for default RLMM is described in detail here. When a partition is deleted, leader updates its state in RLMM with DELETE_MARKED. A task on each leader of “__remote_log_segment_metadata_topic” partitions consume the messages and checks for messages which have the state as “DELETE_MARKED” and schedules them to be deleted. It commits consumer offsets upto which these markers were handled so that whenever a leader switches to other brokers they can continue from where they were left. The controller considers the topic partition is deleted only when it determines that there are no log segments for that topic partition by using RLMM.

RemoteLogMetadataManager implemented with an internal topic

Metadata of remote log segments are stored in an internal non compact topic called `__remote_log_segment_metadata`. This topic can be created with default partitions count as 50. Users can configure the partitions count and replication factor etc as mentioned in the config section.

...

RLMM registers the topic partitions that the broker is either a leader or a follower. . These topic partitions include the remote log metadata topic partitions also.

RLMM maintains metadata cache by subscribing to the respective remote log metadata topic partitions. Whenever a topic partition is reassigned to a new broker and RLMM on that broker is not subscribed to the respective remote log metadata topic partition then it will subscribe to the respective remote log metadata topic partition and adds all the entries to the cache. So, in the worst case, RLMM on a broker may be consuming from most of the remote log metadata topic partitions. This requires the cache to be based on disk storage like RocksDB to avoid a high memory footprint on a brokerIn the initial version, we will have a file-based cache for all the messages that are already consumed by this instance and it will load in-memory whenever RLMM is started. This cache is maintained in a separate file for each of the topic partitions. This will allow us to commit offsets of the partitions that are already read. Committed offsets can be stored in a local file to avoid reading the messages again when a broker is restarted. 

Image Removed

Message Format

RLMM instance on broker publishes the message to the topic with key as null and value with the below format.

type      : unsigned var int, represents the value type. This value is 'apikey' as mentioned in the schema. 
version : unsigned var int, the 'version' number of the type as mentioned in the schema. 
data      : record payload in kafka protocol message format, the schema is given below.

Schema can be evolved by adding a new version with the respective changes. A new type can also be supported by adding with the respective type and its version. 

RLMM segment overhead:

Topic partition's topic-id : uuid : 2 longs.

remoteLogSegmentId : uuid : 2 longs.

remoteLogSegmentMetadata : 5 longs + 1 int +1 byte + ~3 epochs(approx avg)

It has leader epochs in-memory which will be much less.

On avg: 10 longs : 10 * 8 = 80 *(other overhead 1.25) = 100 bytes

When a segment is rolled on a broker per sec.

retention as 30days : 60*60*24*30 ~ 2.6MM

2.6MM segments would take ~ 260MB. (This is 1% in our production env)

This overhead is not that significant as brokers may be using several GBs of memory. 

We can also have a lazy load approach by keeping only minimal in-memory entries like offset, epoch, uuid, and entry position in the file. When it is needed we can access it by using the entry position in the file. 

Message Format

RLMM instance on broker publishes the message to the topic with key as null and value with the below format. 

type      : Represents the value type. This value is 'apikey' as mentioned in the schema. Its type is 'byte'.
version : the 'version' number of the type as mentioned in the schema. Its type is 'byte'.
data      : record payload in kafka protocol message format, the schema is given below.

Both type and version are added before the data is serialized into record value.  Schema can be evolved by adding a new version with the respective changes. A new type can also be supported by adding the respective type and its version.  

{
  "apiKey": 0,
  "type": "data",
  "name": "RemoteLogSegmentMetadataRecord",
  "validVersions": "0",
  "flexibleVersions": "none",
  "fields": [
    {
      "name": "RemoteLogSegmentId",
      "type": "RemoteLogSegmentIdEntry",
      "versions": "0+",
      "about": "Unique idrepresentation of the remote log segment",
      "fields": [
        {
          "name": "topicTopicIdPartition",
          "type": "stringTopicIdPartitionEntry",
          "versions": "0+",
          "about": "Topic name"Represents unique topic partition",
        },   "fields": [
            {
              "name": "partitionName",
              "type": "int32string",
              "versions": "0+",
              "about": "PartitionTopic numbername"
            },
            {
              "name": "idId",
              "type": "uuid",
              "versions": "0+",
              "about": "Unique identifier of the topic"
            },
            ]{
    },
    {
      "name": "StartOffsetPartition",
              "type": "int64int32",
              "versions": "0+",
              "about": "Partition "Start offset  of the segment."
number"
            }
          ]
        },
        {
          "name": "endOffsetId",
          "type": "int64uuid",
          "versions": "0+",
          "about": "EndUnique offset identifier of the remote log segment."
    },
    {}
      ]
    },
    {
      "name": "LeaderEpochStartOffset",
      "type": "int32int64",
      "versions": "0+",
      "about": "LeaderStart epochoffset from whichof thisthe segment instance is created or updated."
    },
    {
      "name": "MaxTimestampEndOffset",
      "type": "int64",
      "versions": "0+",
      "about": "MaximumEnd timestampoffset with inof thisthe segment."
    },
    {
      "name": "EventTimestampLeaderEpoch",
      "type": "int64int32",
      "versions": "0+",
      "about": "EventLeader epoch timestampfrom ofwhich this segment. instance is created or updated"
    },
    {
      "name": "SegmentLeaderEpochsMaxTimestamp",
      "type": "[]SegmentLeaderEpochEntryint64",
      "versions": "0+",
      "about": "EventMaximum timestamp with ofin this segment.",
      "fields": [
 },
       {
          "name": "LeaderEpochEventTimestamp",
          "type": "int32int64",
          "versions": "0+",
          "about": "LeaderEvent epoch"
timestamp of this segment."
     },
        {
          "name": "OffsetSegmentLeaderEpochs",
          "type": "int64[]SegmentLeaderEpochEntry",
          "versions": "0+",
          "about": "StartLeader offset for the leader epoch"
epoch cache.",
           }"fields": [
      ]
    },{
    {
      "name": "SegmentSizeInBytesLeaderEpoch",
          "type": "int64int32",
          "versions": "0+",
          "about": "Segment size in bytes"
Leader epoch"
        },
        {
          "name": "StateOffset",
          "type": "int8int64",
          "versions": "0+",
          "about": "StateStart offset offor the segmentleader epoch"
    }
    ]
}


	/**
     * It]
 indicates the state of},
 the remote log segment.{
 This will be based on the action executed on this segment by
     * remote log service implementation."name": "SegmentSizeInBytes",
      "type": "int32",
      "versions": "0+",
     *
 "about": "Segment size in * todo: check whether the state validations to be checked or not, add next possible states for each state.bytes"
    },
    {
      "name": "RemoteLogSegmentState",
      "type": "int8",
     */
 "versions": "0+",
     public enum"about": "State {

of the remote log segment"
    /**}
  ]
}


{
  "apiKey": 1,
  "type": "data",
 * This state indicates that the segment copying to remote storage is started but not yet finished."name": "RemoteLogSegmentMetadataRecordUpdate",
  "validVersions": "0",
  "flexibleVersions": "none",
  "fields": [
    {
         */"name": "RemoteLogSegmentId",
        COPY_STARTED((byte) 0),
"type": "RemoteLogSegmentIdEntry",
        /**"versions": "0+",
      "about": "Unique representation *of Thisthe stateremote indicates that the segment copying to remote storage is finished.log segment",
      "fields": [
        {
 */
         COPY_FINISHED((byte) 1),

"name": "TopicIdPartition",
        /**
  "type": "TopicIdPartitionEntry",
      * This segment is marked for delete. That means, it is eligible for deletion. This is used when a topic/partition"versions": "0+",
          "about": "Represents unique topic partition",
         * is deleted so that deletion agents can start deleting them as the leader/follower does not exist."fields": [
            {
         */
     "name":   DELETE_MARKED((byte) 2),
"Name",
              "type": "string",
        /**
      "versions": "0+",
   * This state indicates that the segment deletion is started but not yet finished."about": "Topic name"
         */
   },
          DELETE_STARTED((byte) 3),

  {
               /**
"name": "Id",
           *  This state indicates that the segment is deleted successfully.
"type": "uuid",
               */"versions": "0+",
        DELETE_FINISHED((byte) 4);
      "about": "Unique identifier of the ...topic"
    }

Configs

...

Replication factor of the topic

Default: 3

...

No of partitions of the topic

Default: 50

...

Retention of the topic in milli seconds

Default: 365 * 24 * 60 * 60 * 1000  (1 yr)

...

Listener name to be  used to connect to the local broker by RemoteLogMetadataManager implementation on the broker. Respective endpoint address is passed with  "bootstrap.servers" property while invoking RemoteLogMetadataManager#configure(Map<String, ?> props). 

This is used by kafka clients created in RemoteLogMetadataManager implementation.

...

Any other properties should be prefixed with "remote.log.metadata." and these will be passed to RemoteLogMetadataManager#configure(Map<String, ?> props).

For ex: Security configuration to connect to the local broker for the listener name configured are passed with props.

[We will add more details later about how the resultant state for each topic partition is computed ]

Public Interfaces

Compacted topics will not have remote storage support. 

Configs

...

remote.log.storage.enable - Whether to enable remote log storage or not. Valid values are `true` or `false` and the default value is false. This property gives backward compatibility.

remote.log.storage.manager.class.name - This is mandatory if the remote.log.storage.enable is set as true.

remote.log.metadata.manager.class.name(optional) - This is an optional property. If this is not configured, Kafka uses an inbuilt metadata manager backed by an internal topic.

...

(These configs are dependent on remote storage manager implementation)

remote.log.storage.*

...

(These configs are dependent on remote log metadata manager implementation)

remote.log.metadata.*

...

remote.log.manager.thread.pool.size
Remote log thread pool size, which is used in scheduling tasks to copy segments, fetch remote log indexes and clean up remote log segments.

remote.log.manager.task.interval.ms
The interval at which remote log manager runs the scheduled tasks like copy segments, fetch remote log indexes and clean up remote log segments.

remote.log.reader.threads
Remote log reader thread pool size

remote.log.reader.max.pending.tasks
Maximum remote log reader thread pool task queue size. If the task queue is full, broker will stop reading remote log segments.

...

 User can set the desired config for remote.log.storage.enable property while creating a topic but it is not allowed to be updated after the topic is created. Other remote.log.* properties can be modified. We will support flipping remote.log.storage.enable in next versions.

Below retention configs are similar to the log retention. This configuration is used to determine how long the log segments are to be retained in the local storage. Existing log.retention.* are retention configs for the topic partition which includes both local and remote storage. 

local.log.retention.ms
The number of milli seconds to keep the local log segment before it gets deleted. If not set, the value in `remote.log.retention.minutes` is used. If set to -1, no time limit is applied.

local.log.retention.bytes
The maximum size of local log segments that can grow for a partition before it deletes the old segments. There is no default value, but the above time based retention always applies.

        },
            {
              "name": "Partition",
              "type": "int32",
              "versions": "0+",
              "about": "Partition number"
            }
          ]
        },
        {
          "name": "Id",
          "type": "uuid",
          "versions": "0+",
          "about": "Unique identifier of the remote log segment"
        }
      ]
    },
    {
      "name": "LeaderEpoch",
      "type": "int32",
      "versions": "0+",
      "about": "Leader epoch from which this segment instance is created or updated"
    },
    {
      "name": "EventTimestamp",
      "type": "int64",
      "versions": "0+",
      "about": "Event timestamp of this segment."
    },
    {
      "name": "RemoteLogSegmentState",
      "type": "int8",
      "versions": "0+",
      "about": "State of the remote segment"
    }
  ]
}



{
  "apiKey": 2,
  "type": "data",
  "name": "RemotePartitionDeleteMetadataRecord",
  "validVersions": "0",
  "flexibleVersions": "none",
  "fields": [
    {
      "name": "TopicIdPartition",
      "type": "TopicIdPartitionEntry",
      "versions": "0+",
      "about": "Represents unique topic partition",
      "fields": [
        {
          "name": "Name",
          "type": "string",
          "versions": "0+",
          "about": "Topic name"
        },
        {
          "name": "Id",
          "type": "uuid",
          "versions": "0+",
          "about": "Unique identifier of the topic"
        },
        {
          "name": "Partition",
          "type": "int32",
          "versions": "0+",
          "about": "Partition number"
        }
      ]
    },
    {
      "name": "Epoch",
      "type": "int32",
      "versions": "0+",
      "about": "Epoch (controller or leader) from which this event is created. DELETE_PARTITION_MARKED is sent by the controller. DELETE_PARTITION_STARTED and DELETE_PARTITION_FINISHED are sent by remote log metadata topic partition leader."
    },
    {
      "name": "EventTimestamp",
      "type": "int64",
      "versions": "0+",
      "about": "Event timestamp of this segment."
    },
    {
      "name": "RemotePartitionDeleteState",
      "type": "int8",
      "versions": "0+",
      "about": "Deletion state of the remote partition"
    }
  ]
}

package org.apache.kafka.server.log.remote.storage;
...
/**
 * It indicates the deletion state of the remote topic partition. This will be based on the action executed on this
 * partition by the remote log service implementation.
 */
public enum RemotePartitionDeleteState {

    /**
     * This is used when a topic/partition is determined to be deleted by controller.
     * This partition is marked for delete by controller. That means, all its remote log segments are eligible for
     * deletion so that remote partition removers can start deleting them.
     */
    DELETE_PARTITION_MARKED((byte) 0),

    /**
     * This state indicates that the partition deletion is started but not yet finished.
     */
    DELETE_PARTITION_STARTED((byte) 1),

    /**
     * This state indicates that the partition is deleted successfully.
     */
    DELETE_PARTITION_FINISHED((byte) 2);
...
}


package org.apache.kafka.server.log.remote.storage;
...
/**
 * It indicates the state of the remote log segment or partition. This will be based on the action executed on this
 * segment or partition by the remote log service implementation.
 * <p>
 */
public enum RemoteLogSegmentState {

    /**
     * This state indicates that the segment copying to remote storage is started but not yet finished.
     */
    COPY_SEGMENT_STARTED((byte) 0),

    /**
     * This state indicates that the segment copying to remote storage is finished.
     */
    COPY_SEGMENT_FINISHED((byte) 1),

    /**
     * This state indicates that the segment deletion is started but not yet finished.
     */
    DELETE_SEGMENT_STARTED((byte) 2),

    /**
     * This state indicates that the segment is deleted successfully.
     */
    DELETE_SEGMENT_FINISHED((byte) 3),
...
}

Configs

Committed offsets file format

Committed offsets are stored in a local file `_rlmm_committed_offsets` under log dir. This file contains offset entry for each subscribed remote log metadata partition as "<partition-no> <offset>".

0 2022
4 104
2 498

Internal flat-file store format of remote log metadata

RLMM stores the remote log metadata messages and builds materialized instances in a flat-file store for each user topic partition.

<magic><topic-name><topic-id><metadata-topic-offset><sequence-of-serialized-entries>

magic:                  
    unsigned var int, version of this file format.
topic-name:             
    string, topic name.
topic-id:               
    uuid, uuid of topic
metadata-topic-offset:  
    var long, offset of the remote log metadata topic partition upto which this topic partition's remote log metadata is fetched.
serialized-entries:
   sequence of serialized entries defined as below, more types can be added later if needed.

Serialization of entry is done as mentioned below. This is very similar to the message format mentioned earlier for storing into the metadata topic.  

length    : unsigned var int, length of this entry which is sum of sizes of type, version, and data.
type      : unsigned var int, represents the value type. This value is 'apikey' as mentioned in the schema. 
version   : unsigned var int, the 'version' number of the type as mentioned in the schema. 
data      : record payload in kafka protocol message format, the schema is given below.

Both type and version are added before the data is serialized into record value.  Schema can be evolved by adding a new version with the respective changes. A new type can also be supported by adding the respective type and its version.


{
  "apiKey": 0,
  "type": "data",
  "name": "RemoteLogSegmentMetadataRecordStored",
  "validVersions": "0",
  "flexibleVersions": "none",
  "fields": [
    {
      "name": "SegmentId",
      "type": "uuid",
      "versions": "0+",
      "about": "Unique identifier of the log segment"
    },
    {
      "name": "StartOffset",
      "type": "int64",
      "versions": "0+",
      "about": "Start offset  of the segment."
    },
    {
      "name": "EndOffset",
      "type": "int64",
      "versions": "0+",
      "about": "End offset  of the segment."
    },
    {
      "name": "LeaderEpoch",
      "type": "int32",
      "versions": "0+",
      "about": "Leader epoch from which this segment instance is created or updated"
    },
    {
      "name": "MaxTimestamp",
      "type": "int64",
      "versions": "0+",
      "about": "Maximum timestamp with in this segment."
    },
    {
      "name": "EventTimestamp",
      "type": "int64",
      "versions": "0+",
      "about": "Event timestamp of this segment."
    },
    {
      "name": "SegmentLeaderEpochs",
      "type": "[]SegmentLeaderEpochEntry",
      "versions": "0+",
      "about": "Event timestamp of this segment.",
      "fields": [
        {
          "name": "LeaderEpoch",
          "type": "int32",
          "versions": "0+",
          "about": "Leader epoch"
        },
        {
          "name": "Offset",
          "type": "int64",
          "versions": "0+",
          "about": "Start offset for the leader epoch"
        }
      ]
    },
    {
      "name": "SegmentSizeInBytes",
      "type": "int32",
      "versions": "0+",
      "about": "Segment size in bytes"
    },
    {
      "name": "RemoteLogSegmentState",
      "type": "int8",
      "versions": "0+",
      "about": "State of the remote log segment"
    }
  ]
}


{
  "apiKey": 1,
  "type": "data",
  "name": "DeletePartitionStateRecord",
  "validVersions": "0",
  "flexibleVersions": "none",
  "fields": [
    {
      "name": "Epoch",
      "type": "int32",
      "versions": "0+",
      "about": "Epoch (controller or leader) from which this event is created. DELETE_PARTITION_MARKED is sent by the controller. DELETE_PARTITION_STARTED and DELETE_PARTITION_FINISHED are sent by remote log metadata topic partition leader."
    },
    {
      "name": "EventTimestamp",
      "type": "int64",
      "versions": "0+",
      "about": "Event timestamp of this segment."
    },
    {
      "name": "RemotePartitionDeleteState",
      "type": "int8",
      "versions": "0+",
      "about": "Deletion state of the remote partition"
    }
  ]
}

Message Formatter for the internal topic

`org.apache.kafka.server.log.remote.storage.RemoteLogMetadataFormatter` can be used to format messages received from remote log metadata topic by console consumer. Users can pass properties mentioned in the below block with '–property' while running console consumer with this message formatter. The below block explains the format and it may change later. This formatter can be helpful for debugging purposes.

partition:<val><sep>message-offset:<val><sep>type:<RemoteLogSegmentMetadata | RemoteLogSegmentMetadataUpdate | DeletePartitionState><sep>version:<_no_><vs>event-value:<string representation of the event>

val: represents the respective value of the key. 
sep: represents the separator, default value is: ","

partition : Remote log metata topic partition number. This is optional. 
Use print.partition property to print it, default is false

message-offset : Offset of this message in remote log metadata topic. This is optional. 
Use print.message.offset property to print it, default is false

type: Event value type, which can be one of RemoteLogSegmentMetadata, RemoteLogSegmentMetadataUpdate, DeletePartitionState values.

version: Version number of the event value type. This is optional. 
Use print.version property to print it, default is false

Use print.all.event.value.fields to print the string representation of the event which will include all the fields in the data, default property value is false.

Event value can be of any of the types below:

remote-log-segment-id is represented as "{id:<><sep>topicId:<val><sep>topicName:<val><sep>partition:<val>}" in the event value.
topic-id-partition is represented as "{topicId:<val><sep>topicName:<val><sep>partition:<val>}" in the event value.

For RemoteLogSegmentMetadata 
default representation is "{remote-log-segment-id:<val><sep>start-offset:<val><sep>end-offset:<val><sep>leader-epoch:<val><sep>remote-log-segment-state:<COPY_SEGMENT_STARTED | COPY_SEGMENT_FINISHED | DELETE_SEGMENT_STARTED | DELETE_SEGMENT_FINISHED>}"

For RemoteLogSegmentMetadataUpdate
default representation is "{remote-log-segment-id:<val><sep>leader-epoch:<val><sep>remote-log-segment-state:<COPY_SEGMENT_STARTED | COPY_SEGMENT_FINISHED | DELETE_SEGMENT_STARTED | DELETE_SEGMENT_FINISHED>}"

For DeletePartitionState
default representation is "{topic-id-partition:<val><sep>epoch:<val><sep>remote-partition-delete-state:<DELETE_PARTITION_MARKED | DELETE_PARTITION_STARTED | DELETE_PARTITION_FINISHED>

Anchor
topic-deletion topic-deletion

Topic deletion lifecycle

The controller receives a delete request for a topic. It goes through the existing protocol of deletion and it makes all the replicas offline stop taking any fetch requests.  After all the replicas reach the offline state, the controller publishes an event to the RemoteLogMetadataManager(RLMM) by marking the topic as deleted using RemoteLogMetadataManager.updateRemotePartitionDeleteMetadata with the state as RemotePartitionDeleteState#DELETE_PARTITION_MARKED.  With KIP-516, topics are represented with uuid, and topics can be deleted asynchronously. This allows the remote logs can be garbage collected later by publishing the deletion marker into the remote log metadata topic. RLMM is responsible for asynchronously deleting all the remote log segments of a partition after receiving RemotePartitionDeleteState as DELETE_PARTITION_MARKED. 

Image Added

Default RLMM handles the remote partition deletion by using RemotePartitionRemover(RPRM). 

RPRM instance is created on a broker with the leaders of the remote log segment metadata topic partitions. This task is responsible for removing remote storage of the topics marked for deletion. It consumes messages from those partitions remote log metadata partitions and filters the delete partition events which need to be processed. It collects those partitions and executes deletion of the respective segments using RemoteStorageManager. This is done at regular intervals of remote.partition.remover.task.interval.ms (default value of 1hr). It commits the processed offsets of metadata partitions once the deletions are executed successfully. This will also be helpful to handle leader failovers to a different replica so that it can start processing the messages where it left off. 

RemotePartitionRemover(RPRM) processes the request with the following flow as mentioned in the below diagram. 

1. The controller publishes DELETE_PARTITION_MARKED event to say that the partition is marked for deletion. There can be multiple events published when the controller restarts or failover and this event will be deduplicated by RPRM. 
2. RPRM receives the DELETE_PARTITION_MARKED and processes it if it is not yet processed earlier.
3. RPRM publishes an event DELETE_PARTITION_STARTED that indicates the partition deletion has already been started. 
4. RPRM gets all the remote log segments for the partition using RLMM and each of these remote log segments is deleted with the next steps.RLMM subscribes to the local remote log metadata partitions and it will have the segment metadata of all the user topic partitions associated with that remote log metadata partition.
5. Publish DELETE_SEGMENT_STARTED event with the segment id. 
6. RPRM deletes the segment using RSM 
7. Publish DELETE_SEGMENT_FINISHED event with segment id once it is successful. 
8. Publish DELETE_PARTITION_FINISHED once all the segments have been deleted successfully.

Image Added

Protocol Changes

ListOffsets

Currently, it supports the listing of offsets based on the earliest timestamp and the latest timestamp of the complete log. There is no change in the protocol but the new versions will start supporting listing earliest offsets based on the local logs but not only on the complete log including remote log. This protocol will be updated with the changes from KIP-516 but there are no changes required as mentioned earlier. Request and response versions will be bumped to version 7.

Fetch

We are bumping up fetch protocol to handle new error codes, there are no changes in request and response schemas. When a follower tries to fetch records for an offset that does not exist locally then it returns a new error `OFFSET_MOVED_TO_TIERED_STORAGE`. This is explained in detail here. 

OFFSET_MOVED_TO_TIERED_STORAGE - when the requested offset is not available in local storage but it is moved to tiered storage.

Public Interfaces

Compacted topics will not have remote storage support. 

Configs

Remote Storage Manager

`RemoteStorageManager` is an interface to provide the lifecycle of remote log segments and indexes. More details about how we arrived at this interface are discussed in the document. We will provide a simple implementation of RSM to get a better understanding of the APIs. HDFS and S3 implementation are planned to be hosted in external repos and these will not be part of Apache Kafka repo. This is in line with the approach taken for Kafka connectors.

Copying and Deleting APIs are expected to be idempotent, so plugin implementations can retry safely and overwrite any partially copied content, or not failing when content is already deleted.

package org.apache.kafka.server.log.remote.storage;
...
/**
 * RemoteStorageManager provides the lifecycle of remote log segments that includes copy, fetch, and delete from remote
 * storage.
 * <p>
 * Each upload or copy of a segment is initiated with {@link RemoteLogSegmentMetadata} containing {@link RemoteLogSegmentId}
 * which is universally unique even for the same topic partition and offsets.
 * <p>
 * RemoteLogSegmentMetadata is stored in {@link RemoteLogMetadataManager} before and after copy/delete operations on
 * RemoteStorageManager with the respective {@link RemoteLogSegmentState}. {@link RemoteLogMetadataManager} is
 * responsible for storing and fetching metadata about the remote log segments in a strongly consistent manner.
 * This allows RemoteStorageManager to store segments even in eventually consistent manner as the metadata is already
 * stored in a consistent store.
 * <p>
 * All these APIs are still evolving.
 */
@InterfaceStability.Unstable
public interface RemoteStorageManager extends Configurable, Closeable {


    /**
     * Type of the index file.
     */
    enum IndexType {
        /**
         * Represents offset index.
         */
        Offset,

        /**
         * Represents timestamp index.
         */
        Timestamp,

        /**
         * Represents producer snapshot index.
         */
        ProducerSnapshot,

        /**
         * Represents transaction index.
         */
        Transaction,

        /**
         * Represents leader epoch index.
         */
        LeaderEpoch,
    }       

    /**
     * Copies the given {@link LogSegmentData} provided for the given {@code remoteLogSegmentMetadata}. This includes
     * log segment and its auxiliary indexes like offset index, time index, transaction index, leader epoch index, and
     * producer snapshot index.
     * <p>
     * Invoker of this API should always send a unique id as part of {@link RemoteLogSegmentMetadata#remoteLogSegmentId()}
     * even when it retries to invoke this method for the same log segment data.
     * <p>
     * This operation is expected to be idempotent. If a copy operation is retried and there is existing content already written,
     * it should be overwritten, and do not throw {@link RemoteStorageException}
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment.
     * @param logSegmentData           data to be copied to tiered storage.
     * @throws RemoteStorageException if there are any errors in storing the data of the segment.
     */
    void copyLogSegmentData(RemoteLogSegmentMetadata remoteLogSegmentMetadata,
                            LogSegmentData logSegmentData)
            throws RemoteStorageException;

    /**
     * Returns the remote log segment data file/object as InputStream for the given {@link RemoteLogSegmentMetadata}
     * starting from the given startPosition. The stream will end at the end of the remote log segment data file/object.
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment.
     * @param startPosition            start position of log segment to be read, inclusive.
     * @return input stream of the requested log segment data.
     * @throws RemoteStorageException          if there are any errors while fetching the desired segment.
     * @throws RemoteResourceNotFoundException the requested log segment is not found in the remote storage.
     */
    InputStream fetchLogSegment(RemoteLogSegmentMetadata remoteLogSegmentMetadata,
                                int startPosition) throws RemoteStorageException;

    /**
     * Returns the remote log segment data file/object as InputStream for the given {@link RemoteLogSegmentMetadata}
     * starting from the given startPosition. The stream will end at the smaller of endPosition and the end of the
     * remote log segment data file/object.
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment.
     * @param startPosition            start position of log segment to be read, inclusive.
     * @param endPosition              end position of log segment to be read, inclusive.
     * @return input stream of the requested log segment data.
     * @throws RemoteStorageException          if there are any errors while fetching the desired segment.
     * @throws RemoteResourceNotFoundException the requested log segment is not found in the remote storage.
     */
    InputStream fetchLogSegment(RemoteLogSegmentMetadata remoteLogSegmentMetadata,
                                int startPosition,
                                int endPosition) throws RemoteStorageException;

    /**
     * Returns the index for the respective log segment of {@link RemoteLogSegmentMetadata}.
     * <p>
     * If the index is not present (e.g. Transaction index may not exist because segments create prior to
     * version 2.8.0 will not have transaction index associated with them.),
     * throws {@link RemoteResourceNotFoundException}
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment.
     * @param indexType                type of the index to be fetched for the segment.
     * @return input stream of the requested index.
     * @throws RemoteStorageException          if there are any errors while fetching the index.
     * @throws RemoteResourceNotFoundException the requested index is not found in the remote storage.
     * The caller of this function are encouraged to re-create the indexes from the segment
     * as the suggested way of handling this error.
     */
    InputStream fetchIndex(RemoteLogSegmentMetadata remoteLogSegmentMetadata,
                           IndexType indexType) throws RemoteStorageException;

    /**
     * Deletes the resources associated with the given {@code remoteLogSegmentMetadata}. Deletion is considered as
     * successful if this call returns successfully without any errors. It will throw {@link RemoteStorageException} if
     * there are any errors in deleting the file.
     * <p>
     * This operation is expected to be idempotent. If resources are not found, it is not expected to
     * throw {@link RemoteResourceNotFoundException} as it may be already removed from a previous attempt.
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment to be deleted.
     * @throws RemoteStorageException          if there are any storage related errors occurred.
     */
    void deleteLogSegmentData(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throws RemoteStorageException;
  }


package org.apache.kafka.common;
...
public class TopicIdPartition {

    private final UUID topicId;
    private final TopicPartition topicPartition;

    public TopicIdPartition(UUID topicId, TopicPartition topicPartition) {
        Objects.requireNonNull(topicId, "topicId can not be null");
        Objects.requireNonNull(topicPartition, "topicPartition can not be null");
        this.topicId = topicId;
        this.topicPartition = topicPartition;
    }

    public UUID topicId() {
        return topicId;
    }

    public TopicPartition topicPartition() {
        return topicPartition;
    }

...
}


package org.apache.kafka.server.log.remote.storage;
...
/**
 * This represents a universally unique identifier associated to a topic partition's log segment. This will be
 * regenerated for every attempt of copying a specific log segment in {@link RemoteStorageManager#copyLogSegment(RemoteLogSegmentMetadata, LogSegmentData)}.
 */
public class RemoteLogSegmentId implements Comparable<RemoteLogSegmentId>, Serializable {
    private static final long serialVersionUID = 1L;

    private final TopicIdPartition topicIdPartition;
    private final UUID id;

    public RemoteLogSegmentId(TopicIdPartition topicIdPartition, UUID id) {
        this.topicIdPartition = requireNonNull(topicIdPartition);
        this.id = requireNonNull(id);
    }

    /**
     * Returns TopicIdPartition of this remote log segment.
     * 
     * @return
     */
    public TopicIdPartition topicIdPartition() {
        return topicIdPartition;
    }

    /**
     * Returns Universally Unique Id of this remote log segment.
     *
     * @return
     */
    public UUID id() {
        return id;
    }
...
}

package org.apache.kafka.server.log.remote.storage;
...
/**
 * It describes the metadata about the log segment in the remote storage.
 */
public class RemoteLogSegmentMetadata implements Serializable {

    private static final long serialVersionUID = 1L;

    /**
     * Universally unique remote log segment id.
     */
    private final RemoteLogSegmentId remoteLogSegmentId;

    /**
     * Start offset of this segment.
     */
    private final long startOffset;

    /**
     * End offset of this segment.
     */
    private final long endOffset;

    /**
     * Leader epoch of the broker.
     */
    private final int leaderEpoch;

    /**
     * Maximum timestamp in the segment
     */
    private final long maxTimestamp;

    /**
     * Epoch time at which the respective {@link #state} is set.
     */
    private final long eventTimestamp;

    /**
     * LeaderEpoch vs offset for messages with in this segment.
     */
    private final Map<Int, Long> segmentLeaderEpochs;

    /**
     * Size of the segment in bytes.
     */
    private final int segmentSizeInBytes;

    /**
     * It indicates the state in which the action is executed on this segment.
     */
    private final RemoteLogSegmentState state;

    /**
     * @param remoteLogSegmentId  Universally unique remote log segment id.
     * @param startOffset         Start offset of this segment.
     * @param endOffset           End offset of this segment.
     * @param maxTimestamp        Maximum timestamp in this segment
     * @param leaderEpoch         Leader epoch of the broker.
     * @param eventTimestamp      Epoch time at which the remote log segment is copied to the remote tier storage.
     * @param segmentSizeInBytes  Size of this segment in bytes.
     * @param state               State of the respective segment of remoteLogSegmentId.
     * @param segmentLeaderEpochs leader epochs occurred with in this segment
     */
    public RemoteLogSegmentMetadata(RemoteLogSegmentId remoteLogSegmentId, long startOffset, long endOffset,
                                    long maxTimestamp, int leaderEpoch, long eventTimestamp,
                                    int segmentSizeInBytes, RemoteLogSegmentState state, Map<Int, Long> segmentLeaderEpochs) {
        this.remoteLogSegmentId = remoteLogSegmentId;
        this.startOffset = startOffset;
        this.endOffset = endOffset;
        this.leaderEpoch = leaderEpoch;
        this.maxTimestamp = maxTimestamp;
        this.eventTimestamp = eventTimestamp;
        this.segmentLeaderEpochs = segmentLeaderEpochs;
        this.state = state;
        this.segmentSizeInBytes = segmentSizeInBytes;
    }

    /**
     * @return unique id of this segment.
     */
    public RemoteLogSegmentId remoteLogSegmentId() {
        return remoteLogSegmentId;
    }

    /**
     * @return Start offset of this segment(inclusive).
     */
    public long startOffset() {
        return startOffset;
    }

    /**
     * @return End offset of this segment(inclusive).
     */
    public long endOffset() {
        return endOffset;
    }

    /**
     * @return Leader or controller epoch of the broker from where this event occurred.
     */
    public int brokerEpoch() {
        return brokerEpoch;
    }

    /**
     * @return Epoch time at which this evcent is occurred.
     */
    public long eventTimestamp() {
        return eventTimestamp;
    }

    /**
     * @return
     */
    public int segmentSizeInBytes() {
        return segmentSizeInBytes;
    }

    public RemoteLogSegmentState state() {
        return state;
    }

    public long maxTimestamp() {
        return maxTimestamp;
    }

    public Map<Int, Long> segmentLeaderEpochs() {
        return segmentLeaderEpochs;
    }

...
}

package org.apache.kafka.server.log.remote.storage;
...
public class LogSegmentData {

    private final File logSegment;
    private final File offsetIndex;
    private final File timeIndex;
    private final File txnIndex;
    private final File producerIdSnapshotIndex;
    private final ByteBuffer leaderEpochIndex;

    public LogSegmentData(File logSegment, File offsetIndex, File timeIndex, File txnIndex, File producerIdSnapshotIndex,
                          ByteBuffer leaderEpochIndex) {
        this.logSegment = logSegment;
        this.offsetIndex = offsetIndex;
        this.timeIndex = timeIndex;
        this.txnIndex = txnIndex;
        this.producerIdSnapshotIndex = producerIdSnapshotIndex;
        this.leaderEpochIndex = leaderEpochIndex;
    }

    public File logSegment() {
        return logSegment;
    }

    public File offsetIndex() {
        return offsetIndex;
    }

    public File timeIndex() {
        return timeIndex;
    }

    public File txnIndex() {
        return txnIndex;
    }

    public File producerIdSnapshotIndex() {
        return producerIdSnapshotIndex;
    }

    public ByteBuffer leaderEpochIndex() {
        return leaderEpochIndex;
    }
...
}

RemoteLogMetadataManager

`RemoteLogMetadataManager` is an interface to provide the lifecycle of metadata about remote log segments with strongly consistent semantics. There is a default implementation that uses an internal topic. Users can plugin their own implementation if they intend to use another system to store remote log segment metadata.

package org.apache.kafka.server.log.remote.storage;
...
/**
 * This interface provides storing and fetching remote log segment metadata with strongly consistent semantics.
 * <p>
 * This class can be plugged in to Kafka cluster by adding the implementation class as
 * <code>remote.log.metadata.manager.class.name</code> property value. There is an inbuilt implementation backed by
 * topic storage in the local cluster. This is used as the default implementation if
 * remote.log.metadata.manager.class.name is not configured.
 * </p>
 * <p>
 * <code>remote.log.metadata.manager.class.path</code> property is about the class path of the RemoteLogStorageManager
 * implementation. If specified, the RemoteLogStorageManager implementation and its dependent libraries will be loaded
 * by a dedicated classloader which searches this class path before the Kafka broker class path. The syntax of this
 * parameter is same with the standard Java class path string.
 * </p>
 * <p>
 * <code>remote.log.metadata.manager.listener.name</code> property is about listener name of the local broker to which
 * it should get connected if needed by RemoteLogMetadataManager implementation. When this is configured all other
 * required properties can be passed as properties with prefix of 'remote.log.metadata.manager.listener.
 * </p>
 * "cluster.id", "broker.id" and all other properties prefixed with "remote.log.metadata." are passed when
 * {@link #configure(Map)} is invoked on this instance.
 * <p>
 */
@InterfaceStability.Evolving
public interface RemoteLogMetadataManager extends Configurable, Closeable {

    /**
     * Asynchronously adds {@link RemoteLogSegmentMetadata} with the containing {@link RemoteLogSegmentId} into {@link RemoteLogMetadataManager}.
     * <p>
     * RemoteLogSegmentMetadata is identified by RemoteLogSegmentId and it should have the initial state which is {@link RemoteLogSegmentState#COPY_SEGMENT_STARTED}.
     * <p>
     * {@link #updateRemoteLogSegmentMetadata(RemoteLogSegmentMetadataUpdate)} should be used to update an existing RemoteLogSegmentMetadata.
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment.
     * @throws RemoteStorageException   if there are any storage related errors occurred.
     * @throws IllegalArgumentException if the given metadata instance does not have the state as {@link RemoteLogSegmentState#COPY_SEGMENT_STARTED}
     * @return a Future which will complete once this operation is finished.
     */
    Future<Void> addRemoteLogSegmentMetadata(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throws RemoteStorageException;

    /**
     * This method is used to update the {@link RemoteLogSegmentMetadata} asynchronously. Currently, it allows to update with the new
     * state based on the life cycle of the segment. It can go through the below state transitions.
     * <p>
     * <pre>
     * +---------------------+            +----------------------+
     * |COPY_SEGMENT_STARTED |----------->|COPY_SEGMENT_FINISHED |
     * +-------------------+-+            +--+-------------------+
     *                     |                 |
     *                     |                 |
     *                     v                 v
     *                  +--+-----------------+-+
     *                  |DELETE_SEGMENT_STARTED|
     *                  +-----------+----------+
     *                              |
     *                              |
     *                              v
     *                  +-----------+-----------+
     *                  |DELETE_SEGMENT_FINISHED|
     *                  +-----------------------+
     * </pre>
     * <p>
     * {@link RemoteLogSegmentState#COPY_SEGMENT_STARTED} - This state indicates that the segment copying to remote storage is started but not yet finished.
     * {@link RemoteLogSegmentState#COPY_SEGMENT_FINISHED} - This state indicates that the segment copying to remote storage is finished.
     * <br>
     * The leader broker copies the log segments to the remote storage and puts the remote log segment metadata with the
     * state as “COPY_SEGMENT_STARTED” and updates the state as “COPY_SEGMENT_FINISHED” once the copy is successful.
     * <p></p>
     * {@link RemoteLogSegmentState#DELETE_SEGMENT_STARTED} - This state indicates that the segment deletion is started but not yet finished.
     * {@link RemoteLogSegmentState#DELETE_SEGMENT_FINISHED} - This state indicates that the segment is deleted successfully.
     * <br>
     * Leader partitions publish both the above delete segment events when remote log retention is reached for the
     * respective segments. Remote Partition Removers also publish these events when a segment is deleted as part of
     * the remote partition deletion.
     *
     * @param remoteLogSegmentMetadataUpdate update of the remote log segment metadata.
     * @throws RemoteStorageException          if there are any storage related errors occurred.
     * @throws RemoteResourceNotFoundException when there are no resources associated with the given remoteLogSegmentMetadataUpdate.
     * @throws IllegalArgumentException        if the given metadata instance has the state as {@link RemoteLogSegmentState#COPY_SEGMENT_STARTED}
     * @return a Future which will complete once this operation is finished.
     */
    Future<Void> updateRemoteLogSegmentMetadata(RemoteLogSegmentMetadataUpdate remoteLogSegmentMetadataUpdate)
            throws RemoteStorageException;

    /**
     * Returns {@link RemoteLogSegmentMetadata} if it exists for the given topic partition containing the offset with
     * the given leader-epoch for the offset, else returns {@link Optional#empty()}.
     *
     * @param topicIdPartition topic partition
     * @param epochForOffset   leader epoch for the given offset
     * @param offset           offset
     * @return the requested remote log segment metadata if it exists.
     * @throws RemoteStorageException if there are any storage related errors occurred.
     */
    Optional<RemoteLogSegmentMetadata> remoteLogSegmentMetadata(TopicIdPartition topicIdPartition,
                                                                int epochForOffset,
                

Remote Storage Manager

`RemoteStorageManager` is an interface to provide the lifecycle of remote log segments and indexes. More details about how we arrived at this interface are discussed in the document. We will provide a simple implementation of RSM to get a better understanding of the APIs. HDFS and S3 implementation are planned to be hosted in external repos and these will not be part of Apache Kafka repo. This is inline with the approach taken for Kafka connectors.

/**
 * RemoteStorageManager provides the lifecycle of remote log segments that includes copy, fetch, and delete from remote
 * storage.
 * <p>
 * Each upload or copy of a segment is initiated with {@link RemoteLogSegmentMetadata} containing {@link RemoteLogSegmentId}
 * which is universally unique even for the same topic partition and offsets.
 * <p>
 * RemoteLogSegmentMetadata is stored in {@link RemoteLogMetadataManager} before and after copy/delete operations on
 * RemoteStorageManager with the respective {@link RemoteLogSegmentMetadata.State}. {@link RemoteLogMetadataManager} is
 * responsible for storing and fetching metadata about the remote log segments in a strongly consistent manner.
 * This allows RemoteStorageManager to store segments even in eventually consistent manner as the metadata is already
 * stored in a consistent store.
 * <p>
 * All these APIs are still evolving.
 */
@InterfaceStability.Unstable
public interface RemoteStorageManager extends Configurable, Closeable {
    InputStream EMPTY_INPUT_STREAM = new ByteArrayInputStream(new byte[0]);

    /**
     * Copies LogSegmentData provided for the given {@param remoteLogSegmentMetadata}.
     * <p>
     * Invoker of this API should always send a unique id as part of {@link RemoteLogSegmentMetadata#remoteLogSegmentId()#id()}
     * even when it retries to invoke this method for the same log segment data.
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment.
     * @param logSegmentData           data to be copied to tiered storage.
     * @throws RemoteStorageException if there are any errors in storing the data of the segment.
     */
    void copyLogSegment(RemoteLogSegmentMetadata remoteLogSegmentMetadata, LogSegmentDatalong logSegmentDataoffset)
            throws RemoteStorageException;

    /**
     * Returns the remotehighest log segmentoffset dataof file/object as InputStreamtopic partition for the given RemoteLogSegmentMetadata starting
     * from the given startPosition. The stream will end at the smaller of endPosition and the end of the remote log leader epoch in remote storage. This is used by
     * remote log management subsystem to know up to which offset the segments have been copied to remote storage for
     * segmenta given dataleader file/object.epoch.
     *
     * @param topicIdPartition topic partition
     * @param remoteLogSegmentMetadata metadata about leaderEpoch      leader epoch
     * @return the remoterequested highest log offset if segmentexists.
     * @param@throws startPositionRemoteStorageException if there are any storage related errors occurred.
    start position*/
 of log segment toOptional<Long> behighestOffsetForEpoch(TopicIdPartition readtopicIdPartition, inclusive.

          * @param endPosition              end position of log segment to be read, inclusive.
     * @return inputint streamleaderEpoch) of the requested log segment data.throws RemoteStorageException;

    /**
     * @throws RemoteStorageException if there are any errors while fetching the desired segment.This method is used to update the metadata about remote partition delete event asynchronously. Currently, it allows updating the
     */
    InputStream fetchLogSegmentData(RemoteLogSegmentMetadata remoteLogSegmentMetadata,
   state ({@link RemotePartitionDeleteState}) of a topic partition in remote metadata storage. Controller invokes
     * this method with {@link RemotePartitionDeleteMetadata} having state as {@link RemotePartitionDeleteState#DELETE_PARTITION_MARKED}.
     * So, remote partition removers can act on this event to clean the respective Longremote startPosition,log Longsegments endPosition)of throwsthe RemoteStorageException;partition.

     /** <p><br>
     * ReturnsIn the offsetcase of indexdefault forRLMM theimplementation, respectiveremote logpartition segmentremover ofprocesses {@link RemoteLogSegmentMetadataRemotePartitionDeleteState#DELETE_PARTITION_MARKED}.
     * <ul>
     * <li> @paramsends remoteLogSegmentMetadataan metadataevent aboutwith thestate remoteas log segment.{@link RemotePartitionDeleteState#DELETE_PARTITION_STARTED}
     * <li> @returngets inputall streamthe ofremote thelog requestedsegments and offsetdeletes indexthem.
     * @throws<li> RemoteStorageExceptionsends ifan thereevent arewith anystate errorsas while fetching the index.
     */{@link RemotePartitionDeleteState#DELETE_PARTITION_FINISHED} once all the remote log segments are
    InputStream fetchOffsetIndex(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throws RemoteStorageException;

* deleted.
     * </**ul>
     *
   Returns the timestamp* index@param forremotePartitionDeleteMetadata theupdate respectiveon logdelete segmentstate of {@linka RemoteLogSegmentMetadata}partition.
     *
 @throws RemoteStorageException         * @paramif remoteLogSegmentMetadatathere metadataare aboutany thestorage remoterelated logerrors segmentoccurred.
     * @throws RemoteResourceNotFoundException *when @returnthere inputare streamno ofresources theassociated requestedwith the timestampgiven indexremotePartitionDeleteMetadata.
     * @throws@return RemoteStorageExceptiona ifFuture therewhich arewill anycomplete errorsonce whilethis fetchingoperation theis indexfinished.
     */
    InputStreamFuture<Void> fetchTimestampIndexputRemotePartitionDeleteMetadata(RemoteLogSegmentMetadataRemotePartitionDeleteMetadata remoteLogSegmentMetadata) remotePartitionDeleteMetadata)
            throws RemoteStorageException;

    /**
     * Returns the transaction index for all the theremote respective log segment metadata of the {@linkgiven RemoteLogSegmentMetadata}topicIdPartition.
     * <p>
     * @param remoteLogSegmentMetadata metadata about the remote log segment. Remote Partition Removers uses this method to fetch all the segments for a given topic partition, so that they
     * @returncan inputdelete streamthem.
 of the requested  transaction index.*
     * @return @throwsIterator RemoteStorageExceptionof ifall therethe areremote anylog errorssegment whilemetadata fetchingfor the given topic indexpartition.
     */
    default InputStream fetchTransactionIndex(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throwsIterator<RemoteLogSegmentMetadata> RemoteStorageException {listRemoteLogSegments(TopicIdPartition topicIdPartition)
        return EMPTY_INPUT_STREAM;
   throws }RemoteStorageException;

    /**
     * Returns theiterator producerof snapshotremote indexlog forsegment themetadata, the respective log segment ofsorted by {@link RemoteLogSegmentMetadataRemoteLogSegmentMetadata#startOffset()}.
     * in
     * @paramascending remoteLogSegmentMetadataorder metadatawhich aboutcontains the remotegiven logleader segmentepoch.
 This is used by *remote @returnlog inputretention stream of the producer snapshot.management subsystem
     * @throwsto RemoteStorageExceptionfetch ifthe theresegment aremetadata anyfor errorsa whilegiven fetchingleader the indexepoch.
     */
    default InputStream* fetchProducerSnapshotIndex(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throws RemoteStorageException {@param topicIdPartition topic partition
     *   return EMPTY_INPUT_STREAM;
@param leaderEpoch    }

  leader  /**epoch
     * Returns@return theIterator leaderof epochremote indexsegments, forsorted the the respective log segment of {@link RemoteLogSegmentMetadata}by start offset in ascending order.
     */
    Iterator<RemoteLogSegmentMetadata> listRemoteLogSegments(TopicIdPartition topicIdPartition,
  * @param remoteLogSegmentMetadata metadata about the remote log segment.
     * @return input stream of the leader epoch index.
     * @throws RemoteStorageException if there are any errors while fetching the index.
     */
    default InputStream fetchLeaderEpochIndex(RemoteLogSegmentMetadata remoteLogSegmentMetadata)
          int leaderEpoch) throws RemoteStorageException {;

    /**
    return EMPTY_INPUT_STREAM;
    }

    /**
     * Deletes the resources associated with the given {@param remoteLogSegmentMetadata}. Deletion is considered as * This method is invoked only when there are changes in leadership of the topic partitions that this broker is
     * responsible for.
     *
 successful   if this* call@param returnsleaderPartitions successfully without anypartitions errors.that Ithave willbecome throwleaders {@linkon RemoteStorageException}this ifbroker.
     * there are any errors in deleting the file@param followerPartitions partitions that have become followers on this broker.
     */
 <p>
   void onPartitionLeadershipChanges(Set<TopicIdPartition> leaderPartitions,
  * {@link RemoteResourceNotFoundException} is thrown when there are no resources associated with the given
     * {@param remoteLogSegmentMetadata}.
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment to be deleted.Set<TopicIdPartition> followerPartitions);

    /**
     * @throws RemoteResourceNotFoundException if This method is invoked only when the requestedtopic resourcepartitions isare notstopped found
on this broker. This can *happen @throwswhen RemoteStorageExceptiona
     * partition is emigrated to ifother therebroker areor anya storagepartition relatedis errors occurreddeleted.
     */
    void deleteLogSegment(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throws RemoteStorageException;

}



/**
 * This represents a universally unique id associated to a topic partition's log segment. This will be regenerated for
 * every attempt of copying a specific* @param partitions topic partitions that have been stopped.
     */
    void onStopPartitions(Set<TopicIdPartition> partitions);
}
package org.apache.kafka.server.log.remote.storage;
...
/**
 * It describes the metadata about the log segment in {@link RemoteLogStorageManager#copyLogSegment(RemoteLogSegmentId, LogSegmentData)}the remote storage.
 */
public class RemoteLogSegmentIdRemoteLogSegmentMetadataUpdate implements Serializable {

    private TopicPartitionstatic topicPartition;
final long serialVersionUID = private UUID id1L;

    public RemoteLogSegmentId(TopicPartition topicPartition, UUID id) {
        this.topicPartition = requireNonNull(topicPartition);
   /**
     * Universally unique remote log segment id.
     */
    private this.idfinal = requireNonNull(id);
    }RemoteLogSegmentId remoteLogSegmentId;

    public TopicPartition topicPartition() {/**
     * Epoch time returnat topicPartition;
which the respective {@link #state}

 is set.
  public UUID id() {*/
    private final long  return ideventTimestamp;

    }
...
}


/**
   * It describes* theLeader metadataepoch aboutof the logbroker segmentfrom inwhere thethis remoteevent storageoccurred.
     */
public   class RemoteLogSegmentMetadataprivate implementsfinal Serializableint {leaderEpoch;

    /**
     * It indicates the state ofin which the remote log action is executed on this segment. This will be based on the action executed on this segment by
     */
    private final RemoteLogSegmentState state;

    /**
     * @param remoteLogSegmentId  Universally unique remote log servicesegment implementationid.
     *
 @param eventTimestamp   * todo: check whetherEpoch thetime stateat validationswhich tothe beremote checkedlog orsegment not,is addcopied nextto possiblethe statesremote fortier each statestorage.
     */
 @param leaderEpoch   public enum State {

   Leader epoch of the broker /**
from where this event occurred.
     * This@param state  indicates  that the segment copying to remote storage is started but not yetstate finished.
of the remote log segment.
     */
    public RemoteLogSegmentMetadataUpdate(RemoteLogSegmentId remoteLogSegmentId,
     COPY_STARTED(),

        /**
         * This state indicates that the segment copying to remote storage is finished.
        long */eventTimestamp,
        COPY_FINISHED(),

        /**
         * This segment is marked for delete. That means, it is eligible for deletion. This is used when a topic/partition
int leaderEpoch,
          * is deleted so that deletion agents can start deleting them as the leader/follower does not exist.
         */
       RemoteLogSegmentState DELETE_MARKED(),

state) {
         /**this.remoteLogSegmentId = remoteLogSegmentId;
        this.eventTimestamp = *eventTimestamp;
 This state indicates that the segment deletion isthis.leaderEpoch started= butleaderEpoch;
 not yet finished.
     this.state = state;
  */
  }

    public RemoteLogSegmentId DELETE_STARTEDremoteLogSegmentId(), {

        return /**remoteLogSegmentId;
    }

    public long createdTimestamp() *{
 This state indicates that the segment is deletedreturn successfully.eventTimestamp;
    }

    public  */
RemoteLogSegmentState state() {
        return DELETE_FINISHED()state;
    }

    
 private static final longpublic serialVersionUIDint = 1L;

leaderEpoch() {
      /**
  return leaderEpoch;
  * Universally unique remote log segment id.
     */ }
...
}

package org.apache.kafka.server.log.remote.storage;
...
public class RemotePartitionDeleteMetadata {

    private final TopicIdPartition topicPartition;
    private final RemoteLogSegmentIdRemotePartitionDeleteState remoteLogSegmentIdstate;

    /**
private final long eventTimestamp;
  * Start offsetprivate offinal this segment.int epoch;

     */
    private final long startOffset;

public RemotePartitionDeleteMetadata(TopicIdPartition topicPartition, RemotePartitionDeleteState state, long eventTimestamp, int epoch) {
      /**
  Objects.requireNonNull(topicPartition);
   * End offset of this segmentObjects.requireNonNull(state);
     */
   if(state private final long endOffset;

    /**
!= RemotePartitionDeleteState.DELETE_PARTITION_MARKED && state != RemotePartitionDeleteState.DELETE_PARTITION_STARTED
       * Leader epoch of the broker.
     */
&& state !=  private final int leaderEpoch;

RemotePartitionDeleteState.DELETE_PARTITION_FINISHED) {
       /**
     *throw Maximum timestamp in the segment
     */new IllegalArgumentException("state should be one of the delete partition states");
    private  final long maxTimestamp;
}
    /**
    this.topicPartition *= EpochtopicPartition;
 time at which the respective {@link #state} is setthis.
state = state;
   */
    private finalthis.eventTimestamp long= eventTimestamp;

    /**
    this.epoch *= LeaderEpochepoch;
 vs offset for messages}

 with in this segment.
public TopicIdPartition topicPartition() {
  */
    private final Map<Long, Long> segmentLeaderEpochs;
return topicPartition;
    /**}

    public *RemotePartitionDeleteState Size of the segment in bytes.
state() {
        return */state;
    private final long segmentSizeInBytes;}

    /**
public long eventTimestamp() {
  * It indicates the state in which the actionreturn iseventTimestamp;
 executed on this segment.}

    public int epoch() */{
    private   final Statereturn stateepoch;

    }

...
}


package org.apache.kafka.server.log.remote.storage;
...
/**
 * It indicates the deletion state of *the @paramremote remoteLogSegmentIdtopic partition. UniversallyThis uniquewill remotebe logbased segmenton id.
the action executed on this
 * partition by @paramthe startOffsetremote log service implementation.
 * <p>
 */
public enum RemotePartitionDeleteState Start{

 offset of this segment./**
     * This is @paramused endOffsetwhen a topic/partition is deleted by controller.
     End* offsetThis ofpartition thisis segment.
marked for delete by controller. *That @parammeans, maxTimestampall its remote log segments are eligible for
 maximum timestamp in this segment
* deletion so that remote *partition @paramremovers leaderEpochcan start deleting them.
     */
 Leader epoch of the broker. DELETE_PARTITION_MARKED((byte) 0),

     /**
 @param eventTimestamp   * EpochThis timestate atindicates whichthat the remotepartition log segmentdeletion is copiedstarted tobut thenot remoteyet tier storagefinished.
     */
 @param segmentSizeInBytes  size of this segment in bytes.DELETE_PARTITION_STARTED((byte) 1),

    /**
     * @paramThis state indicates that The respective segment of remoteLogSegmentIdthe partition is markeddeleted fro deletionsuccessfully.
     * @param/
 segmentLeaderEpochs leader epochs occurred with in this segment
 DELETE_PARTITION_FINISHED((byte) 2);

    private static */
final Map<Byte, RemotePartitionDeleteState> STATE_TYPES public= RemoteLogSegmentMetadata(RemoteLogSegmentId remoteLogSegmentId, long startOffset, long endOffset,
Collections.unmodifiableMap(
            Arrays.stream(values()).collect(Collectors.toMap(RemotePartitionDeleteState::id, Function.identity())));

    private final byte id;

    RemotePartitionDeleteState(byte id) {
        this.id = id;
    }

  long maxTimestamp, intpublic leaderEpoch, long eventTimestamp,byte id() {
        return id;
    }

    public static RemotePartitionDeleteState forId(byte id) {
        return STATE_TYPES.get(id);
    }
...
}

package org.apache.kafka.server.log.remote.storage;
...
/**
 * longIt segmentSizeInBytes,indicates Statethe state, of Map<Long,the Long>remote segmentLeaderEpochs) {
        this.remoteLogSegmentId = remoteLogSegmentId;
        this.startOffset = startOffset;
        this.endOffset = endOffset;
        this.leaderEpoch = leaderEpoch;
        this.maxTimestamp = maxTimestamp;log segment. This will be based on the action executed on this
 * segment by the remote log service implementation.
 * <p>
 */
public enum RemoteLogSegmentState {

    /**
     * This state indicates that the segment copying to remote storage is started but not yet finished.
     */
   this.eventTimestamp = eventTimestamp; COPY_SEGMENT_STARTED((byte) 0),

    /**
     this.segmentLeaderEpochs* =This segmentLeaderEpochs;
state indicates that the segment copying to remote this.statestorage =is state;finished.
     */
   this.segmentSizeInBytes = segmentSizeInBytes; COPY_SEGMENT_FINISHED((byte) 1),

    }
...
}

public class LogSegmentData {

/**
     * privateThis finalstate Fileindicates logSegment;
that the segment deletion privateis finalstarted Filebut offsetIndex;
not yet finished.
  private final File timeIndex;*/
    private final File txnIndex;DELETE_SEGMENT_STARTED((byte) 2),

    private/**
 final File producerIdSnapshotIndex;
  * This privatestate finalindicates File leaderEpochIndex;

    public LogSegmentData(File logSegment, File offsetIndex, File timeIndex, File txnIndex, File producerIdSnapshotIndex,
that the segment is deleted successfully.
     */
    DELETE_SEGMENT_FINISHED((byte) 3),

    private static final Map<Byte, RemoteLogSegmentState> STATE_TYPES = Collections.unmodifiableMap(
            Arrays.stream(values()).collect(Collectors.toMap(RemoteLogSegmentState::id, Function.identity())));

  File leaderEpochIndex) {
private final byte id;

     this.logSegment = logSegment;RemoteLogSegmentState(byte id) {
        this.offsetIndexid = offsetIndexid;
    }

    public this.timeIndex = timeIndex;byte id() {
        this.txnIndex = txnIndexreturn id;
    }

    public  this.producerIdSnapshotIndex = producerIdSnapshotIndex;static RemoteLogSegmentState forId(byte id) {
        return this.leaderEpochIndex = leaderEpochIndexSTATE_TYPES.get(id);
    }
...
...
}

RemoteLogMetadataManager

`RemoteLogMetadataManager` is an interface to provide the lifecycle of metadata about remote log segments with strongly consistent semantics. There is a default implementation that uses an internal topic. Users can plugin their own implementation if they intend to use another system to store remote log segment metadata.

/**
 * This interface provides storing and fetching remote log segment metadata with strongly consistent semantics.
 * <p>
 * This class can be plugged in to Kafka cluster by adding the implementation class as
 * <code>remote.log.metadata.manager.class.name</code> property value. There is an inbuilt implementation backed by
 * topic storage in the local cluster. This is used as the default implementation if
 * remote.log.metadata.manager.class.name is not configured.
 * </p>
 * <p>
 * <code>remote.log.metadata.manager.class.path</code> property is about the class path of the RemoteLogStorageManager
 * implementation. If specified, the RemoteLogStorageManager implementation and its dependent libraries will be loaded
 * by a dedicated classloader which searches this class path before the Kafka broker class path. The syntax of this
 * parameter is same with the standard Java class path string.
 * </p>
 * <p>
 * <code>remote.log.metadata.manager.listener.name</code> property is about listener name of the local broker to which
 * it should get connected if needed by RemoteLogMetadataManager implementation. When this is configured all other
 * required properties can be passed as properties with prefix of 'remote.log.metadata.manager.listener.
 * </p>
 * "cluster.id", "broker.id" and all the properties prefixed with "remote.log.metadata." are passed when
 * {@link #configure(Map)} is invoked on this instance.
 * <p>
 * <p>
 * <p>
 * All these APIs are still evolving.
 * <p>
 * We may refactor TopicPartition in the below APIs to an abstraction that contains a unique identifier
 * and TopicPartition. This will be done once unique identifier for a topic is introduced with
 * <a href="https://cwiki.apache.org/confluence/display/KAFKA/KIP-516%3A+Topic+Identifiers">KIP-516</a>
 */
@InterfaceStability.Unstable
public interface RemoteLogMetadataManager extends Configurable, Closeable {

    /**
     * Stores RemoteLogSegmentMetadata with the containing RemoteLogSegmentId into RemoteLogMetadataManager.
     * <p>
     * RemoteLogSegmentMetadata is identified by RemoteLogSegmentId.
     *
     * @param remoteLogSegmentMetadata metadata about the remote log segment to be deleted.
     * @throws RemoteStorageException if there are any storage related errors occurred.
     */
    void putRemoteLogSegmentData(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throws RemoteStorageException;

    /**
     * Fetches RemoteLogSegmentMetadata for the given topic partition containing offset and leader-epoch for the offset.
     * <p>
     *
     * @param topicPartition topic partition
     * @param offset         offset
     * @param epochForOffset leader epoch for the given offset
     * @return the requested remote log segment metadata.
     * @throws RemoteStorageException if there are any storage related errors occurred.
     */
    RemoteLogSegmentMetadata remoteLogSegmentMetadata(TopicPartition topicPartition, long offset, int epochForOffset)
            throws RemoteStorageException;

    /**
     * Returns earliest log offset if there are segments in the remote storage for the given topic partition and
     * leader epoch else returns {@link Optional#empty()}.
     *
     * @param topicPartition topic partition
     * @param leaderEpoch    leader epoch
     * @return the earliest log offset if exists.
     */
    Optional<Long> earliestLogOffset(TopicPartition topicPartition, int leaderEpoch) throws RemoteStorageException;

    /**
     * Returns highest log offset of topic partition for the given leader epoch in remote storage. This is used by
     * remote log management subsystem to know upto which offset the segments have been copied to remote storage  for
     * a given leader epoch.
     *
     * @param topicPartition topic partition
     * @param leaderEpoch    leader epoch
     * @return the requested highest log offset if exists.
     * @throws RemoteStorageException if there are any storage related errors occurred.
     */
    Optional<Long> highestLogOffset(TopicPartition topicPartition, int leaderEpoch) throws RemoteStorageException;

    /**
     * Deletes the log segment metadata for the given remoteLogSegmentMetadata.
     *
     * @param remoteLogSegmentMetadata remote log segment metadata to be deleted.
     * @throws RemoteStorageException if there are any storage related errors occurred.
     */
    void deleteRemoteLogSegmentMetadata(RemoteLogSegmentMetadata remoteLogSegmentMetadata) throws RemoteStorageException;

    /**
     * List the remote log segment metadata of the given topicPartition.
     * <p>
     * This is used when a topic partition is deleted, to fetch all the remote log segments for the given topic
     * partition and delete them .
     *
     * @return Iterator of remote segments, sorted by baseOffset in ascending order.
     */
    default Iterator<RemoteLogSegmentMetadata> listRemoteLogSegments(TopicPartition topicPartition) {
        return listRemoteLogSegments(topicPartition, 0);
    }

    /**
     * Returns iterator of remote log segment metadata, sorted by {@link RemoteLogSegmentMetadata#startOffset()} in
     * ascending order which contains the given leader epoch. This is used by remote log retention management subsystem
     * to fetch the segment metadata for a given leader epoch and cleansup based on retention policies.
     *
     * @param topicPartition topic partition
     * @param leaderEpoch    leader epoch
     * @return Iterator of remote segments, sorted by baseOffset in ascending order.
     */
    Iterator<RemoteLogSegmentMetadata> listRemoteLogSegments(TopicPartition topicPartition, long leaderEpoch);

    /**
     * This method is invoked only when there are changes in leadership of the topic partitions that this broker is
     * responsible for.
     *
     * @param leaderPartitions   partitions that have become leaders on this broker.
     * @param followerPartitions partitions that have become followers on this broker.
     */
    void onPartitionLeadershipChanges(Set<TopicPartition> leaderPartitions, Set<TopicPartition> followerPartitions);

    /**
     * This method is invoked only when the given topic partitions are stopped on this broker. This can happen when a
     * partition is emigrated to other broker or a partition is deleted.
     *
     * @param partitions topic partitions which have been stopped.
     */
    void onStopPartitions(Set<TopicPartition> partitions);
}

New Metrics

The following new metrics will be added:

...

kafka.server:type=BrokerTopicMetrics, name=RemoteBytesOutPerSec, topic=([-.w]+)

...

}

New Metrics

The following new metrics will be added:

Some of these metrics have been updated with new names as part of KIP-930

Upgrade

Follow the steps mentioned in Kafka upgrade to reach the state where all brokers are running on the latest binaries with the respective "inter.broker.protocol" and "log.message.format" versions. Tiered storage requires the message format to be > 0.11.

To enable tiered storage subsytems, a rolling restart should be done by enabling "remote.log.storage.system.enable" on all brokers.

You can enable tiered storage by setting “remote.storage.enable” to true on the desired topics. Before enabling tiered storage, you should make sure the producer snapshots are built for all the segments for that topic in all followers. You should wait till the log retention occurs for all the segments so that all the segments have producer snapshots. Because follower replicas for topics with tier storage enabled, need the respective producer snapshot for each segment for reconciling the state as mentioned in the earlier follower fetch protocol section.

Downgrade

Downgrade to earlier versions(> 2.1) is possible but the data available only on remote storage will not be available. There will be a few files that are created in remote index cache directory($log.dir/remote-log-index-cache) and other remote log segment metadata cache files that need to be cleaned up by the user. We may provide a script to cleanup the cache files created by tiered storage.Users have to manually delete the data in remote storage based on the bucket or dir configured with tiered storage.

Limitations

• Once tier storage is enabled for a topic, it can not be disabled. We will add this feature in future versions. One possible workaround is to create a new topic and copy the data from the desired offset and delete the old topic. Another possible work around is to set the log.local.retention.ms same as retention.ms and wait until the local retention catches up until complete log retention. This will make the complete data available locally. After that, set remote.storage.enable as false to disable tiered storage on a topic. 
• Multiple Log dirs on a broker are not supported (JBOD related features).
• Tiered storage is not supported for compacted topics.

Integration and System tests

For integration tests, we use file based(LocalTieredStorage) RemoteStorageManager(RSM) . For system tests, we plan to have a single node HDFS cluster in one of the containers and use HDFS RSM implementation.

Feature Test

Feature test cases and test results are documented in this google spreadsheet.

Performance Test Results

We have tested the performance of the initial implementation of this proposal.

The cluster configuration:

1. 5 brokers
2. 20 CPU cores, 256GB RAM (each broker)
3. 2TB * 22 hard disks in RAID0 (each broker)
4. Hardware RAID card with NV-memory write cache
5. 20Gbps network
6. snappy compression
7. 6300 topic-partitions with 3 replicas
8. remote storage uses HDFS

Each test case is tested under 2 types of workload (acks=all and acks=1)

Test case 1 (Normal case):

Normal traffic as described above.

We can see there is a little overhead when tiered storage is turned on. This is expected, as the brokers have to ship segments to remote storage, and sync the remote segment metadata between brokers. With at-least-once (acks=all) produce, the produce latency is slightly increased when tiered storage is turned on. With acks=1 produce, the produce latency is almost not changed when tiered storage is turned on.

Test case 2 (out-of-sync consumers catching up):

In addition to the normal traffic, 9 out-of-sync consumers consume 180MB/s per broker (or 900MB/s in total) old data.

With tiered storage, the old data is read from HDFS. Without tiered storage, the old data is read from local disk.

Consuming old data has a significant performance impact to acks=all producers. Without tiered storage, the P99 produce latency is almost ~1.5 times. With tiered storage, the performance impact is relatively lower, because remote storage reading does not compete with the local hard disk bandwidth with produce requests.

Consuming old data has little impact to acks=1 producers.

Test case 3 (rebuild broker):

Under the normal traffic, stop a broker, remove all the local data, and rebuild it without replication throttling. This case simulates replacing a broken broker server.

With tiered storage, the rebuilding broker only needs to fetch the latest data that has not been shipped to remote storage. Without tiered storage, the rebuilt broker has to fetch all the data that has not expired from the other brokers. With the same log retention time, tiered storage reduced the rebuilding time by more than 100 times.

Without tiered storage, the rebuilding broker has to read a large amount of data from the local hard disks of the leaders. This competes for page cache and local disk bandwidth with the normal traffic and dramatically increases the acks=all produce latency.

Future work

• Enhance RLMM local file-based cache with RocksDB to avoid loading the whole cache inmemory. 
• Enhance RLMM implementation based on topic based storage pointing to a target Kafka cluster instead of using a system level topic within the cluster.
• Improve default RLMM implementation with a less chatty protocol.
• Support disabling tiered storage for a topic. 
• Add a system level config to enable tiered storage for all the topics in a cluster.
• Recovery mechanism in case of the broker or cluster failure.
  • This is to be done by fetching the remote log metadata from RemoteStorageManager.
• Recovering  from remote log metadata topic partitions truncation
• Extract RPMM as a separate task and allow any RLMM implementation to reuse the task for deletion of remote segments and complete the remote partition deletion.

Alternatives considered

Following alternatives were considered:

1. Replace all local storage with remote storage - Instead of using local storage on Kafka brokers, only remote storage is used for storing log segments and offset index files. While this has the benefits related to reducing the local storage, it has the problem of not leveraging the OS page cache and local disk for efficient latest reads as done in Kafka today. 
2. Implement Kafka API on another store - This is an approach that is taken by some vendors where Kafka API is implemented on a different distributed, scalable storage (example HDFS). Such an option does not leverage Kafka other than API compliance and requires the much riskier option of replacing the entire Kafka cluster with another system.
3. Client directly reads remote log segments from the remote storage - The log segments on the remote storage can be directly read by the client instead of serving it from Kafka broker. This reduces Kafka broker changes and has the benefits of removing an extra hop. However, this bypasses Kafka security completely, increases Kafka client library complexity and footprint, causes compatibility issues to the existing Kafka client libraries, and hence is not considered. 
4. Store all remote segment metadata in remote storage. This approach works with the storage systems that provide strong consistent metadata, such as HDFS, but does not work with S3 and GCS. Frequently calling LIST API on S3 or GCS also incurs huge costs. So, we choose to store metadata in a Kafka topic in the default implementation but allow users to use other methods with their own RLMM implementations.
5. Cache all remote log indexes in local storage. Store remote log segment information in local storage. 

   Image Added

Meeting Notes

23 Mar 2021

(Notes by Kowshik)

•  Discussion:
  • Discussed implementation of highestLogOffset and listAllRemoteLogSegments methods in KIP-405 PR: https://github.com/apache/kafka/pull/10218.
  • Discussed implementation of state transition validation checks in RemoteLogSegmentState and cases where the source state can still be null.
  • Discussed Log layer refactor and the plan to extract the recovery logic out of the Log layer in a separate PR.
• Follow-ups:
  • Satish to look into review comments on https://github.com/apache/kafka/pull/10218. Jun/Kowshik to review the PR whenever it is ready again.
  • Satish to raise PR addressing last batch of review comments on the interface PR: https://github.com/apache/kafka/pull/10173.
  • Kowshik to continue working on recovery logic refactor and Log layer refactor.
  • (Done) Kowshik to update the external facing Log layer refactor proposal doc with details about the recovery logic refactor: https://docs.google.com/document/d/1dQJL4MCwqQJSPmZkVmVzshFZKuFy_bCPtubav4wBfHQ/edit# .

09 Feb 2021

• Notes
  • Discussed the downgrade path, KIP will be updated with that.
  • Discussed the limitation of not allowing disable tiered storage on a topic.
  • All are agreed that KIP is ready for voting.

26 Jan 2021

• Notes
  • Discussed the latest review comments from the mail thread.
    
  • Manikumar will review and provide comments.
    

12 Jan 2021

• Notes
  • Satish discussed the edge cases around upgrade path with KIP-516 updates. Jun clarified on how topic-id is received after IBP is udpated on all brokers.

  • Jun suggested to update the KIP with more details on Remote Partition Remover.

  • RLMM flat file format was discussed and Jun asked to clarify the header section.

  • Kowshik and Jun will provide Log layer refactoring writeup.

15 Dec 2020

• Notes
  • Discussed producer snapshot fix missing in 2.7
  • Satish discussed memory growth due to RLMM cache and it looks to be practically negligible. The proposal is to use inmemory cache and checkpoint that to disk.
  • Satish will update the KIP with Upgrade path.
  • Kowshik and Jun will look into LOg refactoring.

08 Dec 2020

•  Discussion Recording
• Notes

  1. Tiered storage upgrade path dicussion:

  • Details need to be documented in the KIP.
  • Current upgrade path plan is based on IBP bump.
  • Enabling of the remote log components may not mean all topics are eligible for tiering at the same time.
  • Should tiered storage be enabled on all brokers before enabling it on any brokers?
  • Is there any replication path dependency for enabling tiered storage?

  
  

  2. RLMM persistence format:

  • We agreed to document the persistence format for the materialized state of default RLMM implementation (topic-based).
  • (carry over from earlier discussion) For the file-based design, we don't know yet the % of increase in memory, assuming the majority of segments are in remote storage. It will be useful to document an estimation for this.

  
  

  3. Topic deletion lifecycle discussion:

  • Under topic deletion lifecycle, step (4) it would be useful to mention how the RemotePartitionRemover (RPRM) gets the list of segments to be deleted, and whether it has any dependency with the RLMM topic.

  
  

  4. Log layer discussion:

  • We discussed the complexities surrounding making code changes to Log layer (Log.scala).
  • Today, the Log class holds attributes and behavior related with local log. In the future, we would have to change the Log layer such that it would also contain the logic for the tiered portion of the log. This addition can pose a maintenance challenge.
  • Some of the existing attributes in the Log layer such as LeaderEpochCache and ProducerStateManager can be related with global view of the log too (i.e. global log is local log + tiered log). It can be therefore useful to think about preparatory refactoring, to see whether we can separate responsibilities related with the local log from the tiered log, and, perhaps provide a global view of the log that combines together both as and when required. The global view of the log could manage the lifecycle of LeaderEpochCache and ProducerStateManager.
  
  Follow-ups:
  
  • KIP-405 updates (upgrade path, RLMM file format and topic deletion)
  • Log layer changes

  
  (Notes taken by Kowshik)

01 Dec 2020

•  Discussion Recording
• Notes

• Satish discussed KIP-405 updates:

  1. Addressed some of the outstanding review comments from previous weeks.
  2. Remote log manager (RLM) cache configuration was added.
  3. Updated default values in the KIP for certain configuration parameters.
  4. RLMM committed offsets are stored in separate files.
  5. Initial version: go ahead with in-memory RLMM materializer implementation for now. Future switch to RocksDB seems feasible since it is an internal change only to RLMM cache.
  6. Yet to update the KIP with KIP-516 (topic ID) changes.
  7. Tiered storage upgrade path details are a work-in-progress. Will be added to the KIP.
  RLMM cache choice: RocksDB-based vs file-based:
  
  1. Harsha/Satish didn't see significant improvement in performance when they tried RocksDB in their prototype.
  2. Other advantages of RocksDB were discussed - snapshots, tooling, checksums etc.
  3. As for file-based design, we don't know yet the % of increase in memory, assuming the majority of segments are in remote storage.
  4. Currently a single file-based implementation for the whole broker is considered. But this may have some issues, so it can be useful to consider a file per partition.
  5. More details needed to be added to the KIP on file management, metadata operations, persisted data format and estimates on memory usage.
  Topic ID:
  
  1. KIP-516 PR may land by end of the year, so we should be able to use it in KIP-405.
  2. Satish to update KIP with details.
  (Notes taken by Kowshik)

10 Nov 2020

•  Discussion Recording
• Notes

29 Sep 2020

•  Discussion Recording
• Notes
  •  Discussed that we can have producerid snapshot for each log segment which will be copied to remote storage. There is already a PR for KAFKA-9393 which addresses similar requirements.
  • Discussed on a case when the local data is not available on brokers, whether it is possible to recover the state from remote storage. 
  • We will update the KIP by early next week with
    • Topic deletion design proposed/discussed in the earlier meeting. This includes the schemas of remote log segment metadata events stored in the topic. 
    • Producerid snapshot for each segment discussion.
    • ListOffsets API version bump to support offset for the earliest local timestamp.
    • Justifying the rationale behind keeping RLMM and local leader epoch as the source of truth. 
    • Rocks DB instances as cache for remote log segment metadata.
    • Any other missing updates planned earlier. 

15 Sep 2020

•  Discussion Recording
• Notes
  •  Discussed the proposed topic deletion lifecycle with and without KIP-516.
    • We will update the KIP with the design details.

    • Jun mentioned that KIP-516 will be available in 3.0 and we can go with the design assuming TopicId support.

  • Discussed on remote log metadata truncation and losing the data of Kafka brokers local storage. 

    • We will update KIP on possible approaches and add any possible APIs needed for RemtoeStorageManager(low Priority for now). 

Feature Test

Feature test cases and test results are documented in this google spreadsheet.

Performance Test Results

We have tested the performance of the initial implementation of this proposal.

The cluster configuration:

1. 5 brokers
2. 20 CPU cores, 256GB RAM (each broker)
3. 2TB * 22 hard disks in RAID0 (each broker)
4. Hardware RAID card with NV-memory write cache
5. 20Gbps network
6. snappy compression
7. 6300 topic-partitions with 3 replicas
8. remote storage uses HDFS

Each test case is tested under 2 types of workload (acks=all and acks=1)

...

Workload-1

(at-least-once, acks=all)

...

Workload-2

(acks=1)

...

10 producers

30MB / sec / broker (leader)

~62K messages / sec / broker (leader)

...

10 producers

55MB / sec / broker (leader)

~120K messages / sec / broker (leader)

...

10 consumers

120MB / sec / broker

~250K messages / sec / broker

...

10 consumers

220MB / sec / broker

~480K messages / sec / broker

Test case 1 (Normal case):

Normal traffic as described above.

...

Workload-1

(acks=all, low traffic)

...

Workload-2

(acks=1, high traffic)

...

We can see there is a little overhead when tiered storage is turned on. This is expected, as the brokers have to ship segments to remote storage, and sync the remote segment metadata between brokers. With at-least-once (acks=all) produce, the produce latency is slightly increased when tiered storage is turned on. With acks=1 produce, the produce latency is almost not changed when tiered storage is turned on.

Test case 2 (out-of-sync consumers catching up):

In addition to the normal traffic, 9 out-of-sync consumers consume 180MB/s per broker (or 900MB/s in total) old data.

With tiered storage, the old data is read from HDFS. Without tiered storage, the old data is read from local disk.

...

Workload-1

(acks=all, low traffic)

...

Workload-2

(acks=1, high traffic)

...

Consuming old data has a significant performance impact to acks=all producers. Without tiered storage, the P99 produce latency is almost tripled. With tiered storage, the performance impact is relatively lower, because remote storage reading does not compete the local hard disk bandwidth with produce requests.

Consuming old data has little impact to acks=1 producers.

Test case 3 (rebuild broker):

Under the normal traffic, stop a broker, remove all the local data, and rebuild it without replication throttling. This case simulates replacing a broken broker server.

...

Workload-1

(acks=all,

12TB data per broker)

...

Workload-2

(acks=1,

34TB data per broker)

...

With tiered storage, the rebuilding broker only needs to fetch the latest data that has not been shipped to remote storage. Without tiered storage, the rebuilt broker has to fetch all the data that has not expired from the other brokers. With the same log retention time, tiered storage reduced the rebuilding time by more than 100 times.

Without tiered storage, the rebuilding broker has to read a large amount of data from the local hard disks of the leaders. This competes page cache and local disk bandwidth with the normal traffic, and dramatically increases the acks=all produce latency.

Future work

• Enhance RLMM implementation based on topic based storage pointing to a target Kafka cluster instead of using as system level topic with in the cluster.  
• Improve default RLMM implementation with less chatty protocol.

Alternatives considered

Following alternatives were considered:

1. Replace all local storage with remote storage - Instead of using local storage on Kafka brokers, only remote storage is used for storing log segments and offset index files. While this has the benefits related to reducing the local storage, it has the problem of not leveraging the OS page cache and local disk for efficient latest reads as done in Kafka today. 
2. Implement Kafka API on another store - This is an approach that is taken by some vendors where Kafka API is implemented on a different distributed, scalable storage (example HDFS). Such an option does not leverage Kafka other than API compliance and requires the much riskier option of replacing the entire Kafka cluster with another system.
3. Client directly reads remote log segments from the remote storage - The log segments on the remote storage can be directly read by the client instead of serving it from Kafka broker. This reduces Kafka broker changes and has benefits of removing an extra hop. However, this bypasses Kafka security completely, increases Kafka client library complexity and footprint, causes compatibility issues to the existing Kafka client libraries, and hence is not considered. 
4. Store all remote segment metadata in remote storage. This approach works with the storage systems that provide strong consistent metadata, such as HDFS, but does not work with S3 and GCS. Frequently calling LIST API on S3 or GCS also incurs huge costs. So, we choose to store metadata in a Kafka topic in the default implementation, but allow users to use other methods with their own RLMM implementations.
5. Cache all remote log indexes in local storage. Store remote log segment information in local storage. 

   Image Removed

...

01 Sep 2020

•  Discussion Recording 
• Notes
  • Topic deletion lifecycle
    • Have a separate section
    • Discuss handling deletions when there is no leader. 
    • Describe the approaches with and without KIP-516 support.
  • Describe more on how are duplicate log segments in remote storage are handled. This is partly covered in example scenarios but good to describe them in the details section. 
  • Discuss more on remote log segment metadata topic truncation. 
  • Remote log segment metadata topic event format
    • the event change log approach instead of having an effective event as a message.
  • Behaviour of APIs with remote storage errors.

...

• Discussion Recording
• Notes
  • KIP is updated with follower fetch protocol and ready to reviewed
  • Satish to capture schema of internal metadata topic in the KIP
  • We will update the KIP with details of different cases
  • Test plan will be captured in a doc and will add to the KIP
  • Add a section "Limitations" to capture the capabilities that will be introduced with this KIP and what will not be covered in this KIP.

Other associated KIPs

KIP-852: Optimize calculation of size for log in remote tier

KIP-917: Additional custom metadata for remote log segment