如何用Bash脚本自动关闭本地Kafka集群？

核心问题原因 #

你记录的PID是启动脚本（如kafka-server-start.sh）的父进程ID，但这类脚本会fork出JVM子进程实际运行Kafka服务。仅杀死父进程不会终止子进程，导致集群无法彻底关闭，这就是你看到多个关联PID的原因。

方法一：使用Kafka官方停止脚本（最可靠） #

Kafka自带配套的停止脚本，能自动识别对应配置的进程并终止，无需手动管理PID。

关闭脚本代码 #

#!/bin/bash

KAFKA_SERVER_DIR="./kafka-3.7.0-src"
cd "$KAFKA_SERVER_DIR" || exit

# 关闭两个Kafka Broker
echo "Stopping Kafka brokers..."
bin/kafka-server-stop.sh config/server.properties
bin/kafka-server-stop.sh config/server1.properties

# 关闭Zookeeper
echo "Stopping Zookeeper..."
bin/zookeeper-server-stop.sh config/zookeeper.properties

# 可选：清理集群临时数据目录
rm -rf /tmp/kafka-logs /tmp/kafka-logs1 /tmp/zookeeper /tmp/kraft-combined-logs

方法二：改进启动脚本，记录进程组ID并批量终止 #

如果需要自定义进程管理，可记录每个服务的进程组ID，通过杀死整个进程组来彻底终止所有关联进程。

改进后的启动脚本 #

#!/bin/bash

KAFKA_SERVER_DIR="./kafka-3.7.0-src"
cd "$KAFKA_SERVER_DIR" || exit
PID_FILE="kafka_pids.txt"
PGID_FILE="kafka_pgids.txt"

# 清理旧数据和记录文件
rm -rf /tmp/kafka-logs /tmp/kafka-logs1 /tmp/zookeeper /tmp/kraft-combined-logs
rm -f "$PID_FILE" "$PGID_FILE"

# 启动Zookeeper并记录进程组ID
start_zookeeper() {
    echo "Starting Kafka Zookeeper..."
    bin/zookeeper-server-start.sh config/zookeeper.properties
}
start_zookeeper &
ZK_PID=$!
ZK_PGID=$(ps -o pgid= -p $ZK_PID | grep -o '[0-9]*')
echo "$ZK_PID $ZK_PGID" >> "$PID_FILE"
echo "$ZK_PGID" >> "$PGID_FILE"
sleep 1

# 启动第一个Broker
create_broker() {
    local config_file="$1"
    echo "Creating Kafka broker with config $config_file "
    bin/kafka-server-start.sh "$config_file"
}
create_broker "config/server.properties" &
BROKER1_PID=$!
BROKER1_PGID=$(ps -o pgid= -p $BROKER1_PID | grep -o '[0-9]*')
echo "$BROKER1_PID $BROKER1_PGID" >> "$PID_FILE"
echo "$BROKER1_PGID" >> "$PGID_FILE"
sleep 2

# 启动第二个Broker
create_broker "config/server1.properties" &
BROKER2_PID=$!
BROKER2_PGID=$(ps -o pgid= -p $BROKER2_PID | grep -o '[0-9]*')
echo "$BROKER2_PID $BROKER2_PGID" >> "$PID_FILE"
echo "$BROKER2_PGID" >> "$PGID_FILE"
sleep 2

# 创建Topic（该进程执行后自动退出，无需记录）
make_topic() {
    local topic_name="$1"
    local replication_factor="$2"
    local num_partitions="$3"
    bin/kafka-topics.sh --create --topic "$topic_name" --bootstrap-server localhost:9092 --replication-factor $replication_factor --partitions $num_partitions
}
make_topic "test_topic" 2 3 > make_topic_error.log &

对应的关闭脚本 #

#!/bin/bash

KAFKA_SERVER_DIR="./kafka-3.7.0-src"
cd "$KAFKA_SERVER_DIR" || exit
PGID_FILE="kafka_pgids.txt"

if [ -f "$PGID_FILE" ]; then
    echo "Stopping Kafka cluster..."
    # 遍历进程组ID，先发送终止信号，再强制清理残留进程
    while read -r PGID; do
        if [ -n "$PGID" ]; then
            echo "Killing process group $PGID..."
            kill -TERM -- -$PGID
            sleep 1
            kill -KILL -- -$PGID 2>/dev/null
        fi
    done < "$PGID_FILE"
    # 清理记录文件和临时数据
    rm -f "$PGID_FILE" "$PID_FILE"
    rm -rf /tmp/kafka-logs /tmp/kafka-logs1 /tmp/zookeeper /tmp/kraft-combined-logs
else
    echo "No running Kafka cluster detected (missing PGID file)."
fi

方法三：通过端口查找进程并终止 #

如果不想修改启动脚本，可根据服务端口直接查找并杀死进程（需确保端口未被其他进程占用）：

#!/bin/bash

# 关闭Zookeeper（默认端口2181）
echo "Stopping Zookeeper..."
lsof -i :2181 | grep LISTEN | awk '{print $2}' | xargs kill -TERM
sleep 1
lsof -i :2181 | grep LISTEN | awk '{print $2}' | xargs kill -KILL 2>/dev/null

# 关闭Kafka Broker（默认端口9092、9093）
echo "Stopping Kafka brokers..."
lsof -i :9092,:9093 | grep LISTEN | awk '{print $2}' | xargs kill -TERM
sleep 1
lsof -i :9092,:9093 | grep LISTEN | awk '{print $2}' | xargs kill -KILL 2>/dev/null

# 清理临时数据
rm -rf /tmp/kafka-logs /tmp/kafka-logs1 /tmp/zookeeper /tmp/kraft-combined-logs

内容的提问来源于stack exchange，提问作者SarthakSin