我检查了集群状态，发现在yarn中有60个挂起或正在进行的map/reduce作业。我在yarn中杀死了所有templetoncontroller作业，所有挂起的作业都很快成功完成。我知道templatoncontroller是一个执行内部map/reduce作业的父进程，那么为什么不杀死它们就不会导致其他作业失败呢？

CREATE DATABASE IF NOT EXISTS crnw_db;
USE crnw_db;

CREATE EXTERNAL TABLE IF NOT EXISTS crnw (
Time TIMESTAMP,
Sensorvalue DOUBLE
)
PARTITIONED BY (year STRING, month STRING, day STRING, customer STRING, site STRING, machine STRING, sensorname STRING)
ROW FORMAT DELIMITED FIELDS TERMINATED BY ','
STORED AS TEXTFILE
LOCATION '/data/storage/';
MSCK REPAIR TABLE crnw;