84
votes

I am writing a shell script to monitor kafka brokers.

I have gone through some links and found that if ZooKeeper contains a list of brokers, and if, in this list, the IP address is present, then a kafka broker is running.

I want a command that I can use in my shell script to get the broker list and check whether kafka is running.

Is there any curl command to get the kafka cluster status like elasticsearch?

9
I know that inside the zookeper-shell script in Kafka's bin folder you can call ls /brokers/ids to get the ids of the brokers currently alive. I don't know how to pass that as a parameter to the script though...jimijazz

9 Answers

131
votes

This command will give you the list of the active brokers between brackets:

./bin/zookeeper-shell.sh localhost:2181 ls /brokers/ids
71
votes

Alternate way using Zk-Client:

If you do not prefer to pass arguments to ./zookeeper-shell.sh and want to see the broker details from Zookeeper CLI, you need to install standalone Zookeeper (As traditional Kafka do not comes up with Jline JAR).

Once you install(unzip) the standalone Zookeeper,then:

  • Run the Zookeeper CLI:
    $ zookeeper/bin/zkCli.sh -server localhost:2181 #Make sure your Broker is already running

  • If it is successful, you can see the Zk client running as:

WATCHER::

WatchedEvent state:SyncConnected type:None path:null
[zk: localhost:2181(CONNECTED) 0]
  • From here you can explore the broker details using various commands:

$ ls /brokers/ids # Gives the list of active brokers
$ ls /brokers/topics #Gives the list of topics
$ get /brokers/ids/0 #Gives more detailed information of the broker id '0'

30
votes
echo dump | nc localhost 2181 | grep brokers

(replace localhost with the host where zookeeper is running)

8
votes

Here are a couple of quick functions I use when bash scripting Kafka Data Load into Demo Environments. In this example I use HDP with no security, but it is easily modified to other environments and intended to be quick and functional rather than particularly robust.

The first retrieves the address of the first ZooKeeper node from the config:

ZKS1=$(cat /usr/hdp/current/zookeeper-client/conf/zoo.cfg | grep server.1)
[[ ${ZKS1} =~ server.1=(.*?):[0-9]*:[0-9]* ]]
export ZKADDR=${BASH_REMATCH[1]}:2181
echo "using ZooKeeper Server $ZKADDR"

The second retrieves the Broker IDs from ZooKeeper:

echo "Fetching list of Kafka Brokers"
export BROKERIDS=$(/usr/hdp/current/kafka-broker/bin/zookeeper-shell.sh ${ZKADDR} <<< 'ls /brokers/ids' | tail -1)
export BROKERIDS=${BROKERIDS//[!0-9 ]/}
echo "Found Kafka Broker IDS: $BROKERIDS"

The third parses ZooKeeper again to retrieve the list of Kafka Brokers Host:port ready for use in the command-line client:

unset BROKERS
for i in $BROKERIDS
do
DETAIL=$(/usr/hdp/current/kafka-broker/bin/zookeeper-shell.sh ${ZKADDR} <<< "get /brokers/ids/$i")
[[ $DETAIL =~ PLAINTEXT:\/\/(.*?)\"\] ]]
if [ -z ${BROKERS+x} ]; then BROKERS=${BASH_REMATCH[1]}; else 
BROKERS="${BROKERS},${BASH_REMATCH[1]}"; fi
done
echo "Found Brokerlist: $BROKERS"
5
votes

If you are using new version of Kafka e.g. 5.3.3, you can use

kafka-broker-api-versions  --bootstrap-server BROKER | grep 9092

You just need to pass one of the brokers

2
votes

To use zookeeper commands with shell script try

zookeeper/bin/zkCli.sh -server localhost:2181 <<< "ls /brokers/ids" | tail -n 1. The last line usually has the response details

2
votes

On MacOS, can try:

brew tap let-us-go/zkcli
brew install zkcli

zkcli ls /brokers/ids
zkcli get /brokers/ids/1
0
votes

I did it like this

#!/bin/bash

ZK_HOST="localhost"
ZK_PORT=2181


for i in `echo dump | nc $ZK_HOST $ZK_PORT | grep brokers`
do
    echo $i
    DETAIL=`zkCli -server "$ZK_HOST:$ZK_PORT" get $i 2>/dev/null | tail -n 1`
    echo $DETAIL
done
0
votes

Using Confluent's REST Proxy API v3:

curl -X GET -H "Accept: application/vnd.api+json" localhost:8082/v3/clusters

where localhost:8082 is Kafka Proxy address.