Description:

topicctl get partitions currently displays the partitions status information as below:

Out-of-sync: This mean either under-replicated or offline partition Wrong Leader: This means partition does not have preferred leader replica Ok: This means partition is in a good state (inSync)

We need to know whether a partition is ok (inSync), under-replicated, offline and also check if the partition has a preferred replica leader.

Current topicctl get partitions problems/gaps:

A partition can be under-replicated(out-of-sync) and also Wrong Leader. current topicctl just shows out-of-sync
- Status should be partition status - [ok, offline, under-replicated] -> refer Status column
- Preferred leader should be - [ok, wrong-leader] -> refer Leader column
If a partition is offline, get partitions is showing partition as ok. This is wrong
get partitions can display information for only one topic. There is a need to display all partitions for all/specific topics
Summary for partitions to know just the partition IDs along with status count
Ability to filter partitions by parition status (Allowed: Ok, Offline, Under-replicated)

`topicctl get partitions` Expectations:

This PR will add following features to topicctl

topicctl get partitions -> gets all topics partitions information
topicctl get partitions --status <ok|offline|under-replicated> -> gets all topics partitions information for specified status
topicctl get partitions --summary -> Get the summary of all topics partitions status
topicctl get partitions should take 0 or any number of topic arguments. (Before, get partitions was taking only 1 topic argument)
repl> get partitions <topic> [--summary] (refer below for repl expectations)

`topicctl:repl> get partitions` Expectations:

From terminal, topicctl get partitions can take more than one argument.

From repl, filtering multiple topics can get tricky. Current repl implementation takes only fixed number of words (command.args).

Hence in repl, we will make get partitions work with only one argument (topic) and PartitionStatus as "" implying all status

repl get partitions expect minimum of 3 arguments and maximum of 4
repl> get partitions [topic] -> this works
repl> get partitions [topic] --summary -> this works
repl> get partitions [topic1] [topic2] -> this works only for [topic1]
repl> get partitions [topic1] [topic2] --summary -> this will not work

Partition Status:

Under replicated: partition is in under replicated if the number of ISR are less than the Replicas available for the partition
Offline: partition is in offline state if partition leader broker ID is not found. (i.e) kafka-go metadata call with partitions having ListenerNotFound Error observed for leader partition
Ok: partition does not have under-replicated or offline

Preferred Leader:

Preferred Leader Replica: if the leader partition broker id is equal to first available Replicas broker id
Un Preferred Leader Replica: if the leader partition broker id is not equal to first available Replicas broker id

Considerations:

We leverage kafka-go metadata for all the above partitions statuses
Unlike sarama which gives the offline leader broker id as -1, kafka-go gives the offline broker id as 0 (also Port is 0 with Partition Error information etc)
To make readability better for users, we will modify broker ID from:
- any ISR broker ID that does not have valid Host or Port from 0 to -1
- any Replica broker ID that does not have valid Host or Port from 0 to -1
- (Offline) Leader Broker ID that does not have a valid Host or Port from 0 to -1

Local Set UP Details:

Set up type: Docker
Number of Kafka Nodes: 3 (one in each rack). Broker IDs - 0,1,2 running on ports 19092, 29092, 39092. Kafka containers are kafka-1 (broker ID-0), kafka-2(broker ID-1), kafka-3(broker ID-2)
Number of zookeeper Nodes: 1, running on port 12181
A Mock Producer produces to topic: test-1
A Mock Consumer reads from topic: test-1
topic: __consumer_offsets - partitions: 50, replication: 3, strategy - any
topic: test-1 - partitions 1, replication 3, strategy - any
topic: threepartition - partitions 3, replication 2, strategy - Fixed placement. We will have partitions 0,1 spread in brokers 1,2, partition 2 in brokers 0,1

Build

Build to run locally - go build -o ./build/topicctl -a ./cmd/topicctl
- this is to check if we are able to achieve same results leveraging --broker-addr
- usage: topicctl get partitions --broker-addr localhost:19092
Build to run inside container GOOS=linux GOARCH=amd64 go build -o ./build/topicctl-linux-amd64 -a ./cmd/topicctl
- this is to check if we are able to achieve same results leveraging --zk-addr
- copy to zk container: docker cp ./build/topicctl-linux-amd64 zookeeper:/tmp
usage: docker exec -u root -it zookeeper /tmp/topicctl-linux-amd64 get partitions --zk-addr localhost:2181

topicctl get partitions --help

# ./build/topicctl get partitions --help

Partitions Information

Created a topic threepartition

topic: threepartition

partitions 0,1 replicas are in brokers 1, 2
partition 2 replicas are in brokers 1, 0

Get all partitions status

# ./build/topicctl get partitions --broker-addr localhost:19092 # docker exec -u root -it zookeeper /tmp/topicctl-linux-amd64 get partitions --zk-addr localhost:2181

Get all partitions status summary

# ./build/topicctl get partitions --broker-addr localhost:19092 --summary # docker exec -u root -it zookeeper /tmp/topicctl-linux-amd64 get partitions --zk-addr localhost:2181 --summary

Kill Broker IDs 1, 2

# docker stop kafka-2; docker stop kafka-3

Getting current all partition status

# ./build/topicctl get partitions --broker-addr localhost:19092 # docker exec -u root -it zookeeper /tmp/topicctl-linux-amd64 get partitions --zk-addr localhost:2181

Get current all partitions status --summary

# ./build/topicctl get partitions --broker-addr localhost:19092 --summary # docker exec -u root -it zookeeper /tmp/topicctl-linux-amd64 get partitions --zk-addr localhost:2181 --summary

start broker ID 1

docker start kafka-2

Getting current all partition status

# ./build/topicctl get partitions --broker-addr localhost:19092

repl and get partitions info for `topic: __consumer_offsets`

# ./build/topicctl repl --broker-addr localhost:19092 # get partitions __consumer_offsets # get partitions __consumer_offsets --summary

NOTE: repl can take only one topic argument and optional flag --summary

start broker ID 2

# docker start kafka-3

Getting current all partition status and filter for test-1, threepartition

# ./build/topicctl get partitions --broker-addr localhost:29092 # ./build/topicctl get partitions test-1 threepartition --broker-addr localhost:29092 --summary # ./build/topicctl get partitions test-1 threepartition --broker-addr localhost:29092 --summary --status under-replicated

Getting stable all partition status

# ./build/topicctl get partitions --broker-addr localhost:29092

running preferred leader for topic: partitionthree fixes the wrong leader

Before preferred leader: # ./build/topicctl get partitions threepartition --broker-addr localhost:29092

After preferred leader: # ./build/topicctl get partitions threepartiton --broker-addr localhost:29092

threepartition summary: # ./build/topicctl get partitions threepartition --broker-addr localhost:29092 --summary

threepartition topic summary filter for status ok: # ./build/topicctl get partitions threepartition --broker-addr localhost:29092 --summary --status ok

fetching partition information for topic that does not displays error message

# ./build/topicctl get partitions topicdoesnotexist --broker-addr localhost:29092

segmentio / topicctl