Contents
  1. 1. shell直接写文件
  2. 2. 部署3.x
  3. 3. 文件保护删除
    1. 3.1. 操作方式
    2. 3.2. 配置权限
  4. 4. hadoop-policy.xml
  5. 5. yarn-site
  6. 6. 跨集群复制
    1. 6.1. 目录容量限制
    2. 6.2. webhdfs使用
    3. 6.3. 机架感知
    4. 6.4. 小文件合并
    5. 6.5. 短路读
    6. 6.6. 外置客户端配置withKerberos
  7. 7. 查看文件大小并排序
  8. 8. 修改副本数
  9. 9. HAR
    1. 9.1. 生成
    2. 9.2. 读取
    3. 9.3. 释放
  10. 10. distcp
    1. 10.1. 指定队列
    2. 10.2. 对拷sftp
    3. 10.3. 同时访问两个HA集群
  11. 11. 使用{}读取多个枚举值路径
  12. 12. 数据丢失处理
  13. 13. balance
  14. 14. 清空回收站
  15. 15. Web页面安全
    1. 15.1. 通过nginx配置用户名密码
  16. 16. 数据权限限制
    1. 16.1. 使用者操作方式
    2. 16.2. 配置权限
  17. 17. GroupMapping配置
  18. 18. HDFS文件清理巡检
  19. 19. 查询在读写的hive表
  20. 20. 文件使用检测-审计
    1. 20.1. 所有访问IP
    2. 20.2. 远程拉取的文件是否在用
    3. 20.3. create的文件是否在用
    4. 20.4. 表名前缀分组统计
  21. 21. 节点管理-退役
    1. 21.1. 配置方式1: Hostname-only
    2. 21.2. 配置方式2: Json
  22. 22. 主备切换
  23. 23. 无须重启 刷新配置

shell直接写文件

1
2
dfs dfs -appendToFile - HDFSfile
# 按ctrl+C结束写入

部署3.x

1
2
export JAVA_HOME=/opt/soft/jdk
export HADOOP_PID_DIR=${HADOOP_HOME}
1
2
3
4
<property>
<name>fs.defaultFS</name>
<value>hdfs://localhost:9000</value>
</property>
1
2
3
4
5
6
7
8
9
10
11
12
<property>
<name>dfs.namenode.name.dir</name>
<value>/opt/soft/data/hdfs/name</value>
</property>
<property>
<name>dfs.datanode.data.dir</name>
<value>/opt/soft/data/hdfs/data</value>
</property>
<property>
<name>dfs.replication</name>
<value>1</value>
</property>
1
2
3
4
5
6
7
8
<property>
<name>yarn.nodemanager.aux-services</name>
<value>mapreduce_shuffle</value>
</property>
<property>
<name>yarn.nodemanager.env-whitelist</name>
<value>JAVA_HOME,HADOOP_COMMON_HOME,HADOOP_HDFS_HOME,HADOOP_CONF_DIR,CLASSPATH_PREPEND_DISTCACHE,HADOOP_YARN_HOME,HADOOP_HOME,PATH,LANG,TZ,HADOOP_MAPRED_HOME</value>
</property>
1
2
3
4
5
6
7
8
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
<property>
<name>mapreduce.application.classpath</name>
<value>$HADOOP_MAPRED_HOME/share/hadoop/mapreduce/*:$HADOOP_MAPRED_HOME/share/hadoop/mapreduce/lib/*</value>
</property>
1
2
bin/hdfs namenode -format
sbin/start-dfs.sh

文件保护删除

hdfs删除逻辑为,路径上所有需要删除的文件节点的父目录有w权限

即:即使A用户是根目录的所有者,只要A创建/允许创建了B用户的目录且该目录下有数据,只要A不是超级用户则不能删除。注意若该目录下无文件,则该目录依然能被删除。

操作方式

启动:使用hdfs用户启动hadoop服务

任务部署:ds, 上传jar包、运行任务都将以hadoop运行,生产是数据hadoop写出为755

个人调试/查看数据:可以本地执行hadoop/spark/flink命令执行

数据保护:定期修改关键数据为hdfs用户下755,只有hdfs可删除文件

hdfs dfs -chown hdfs xxx
#hdfs dfs -chmod -R 755 xxx

配置权限

hdfs-site.xml

1
2
3
4
5
6
7
8
<property>
<name>dfs.permissions.enabled</name>
<value>true</value>
</property>
<property>
<name>dfs.namenode.acls.enabled</name>
<value>true</value>
</property>

单个用户设置权限
增加ACL: hdfs dfs -setfacl -m user:jiangmanhua:rwx /hive/warehouse/label.db
递归增加ACL: hdfs dfs -setfacl -R -m user:hadoop:r-x /dir
删除ACL: hdfs dfs -setfacl -x user:hadoop /file
查询ACL: hdfs dfs -getfacl /hive/warehouse/label.db/
hdfs dfs -ls /hive/warehouse/label.db/

hadoop-policy.xml

yarn rmadmin -refreshServiceAcl
hadoop dfsadmin -refreshServiceAcl

基础开关
hadoop.security.authorization=true

控制提交任务/ JobTracker
security.job.client.protocol.acl

访问HDFS /NameNode
security.client.protocol.acl

yarn-site

开启yarn ACLs:
hadoop: core-site.xml
hadoop.security.authorization=true #开启服务级别验证,否则hadoop组件的acl设置不生效
yarn: yarn-site.xml

yarn.acl.enable
true


yarn.admin.acl
hdfs

$ vi $HADOOP_CONF_DIR/capacity-scheduler.xml
$ yarn rmadmin -refreshQueues

yarn.scheduler.capacity.root.<queue-path>.acl_submit_applications

跨集群复制

[Test case from new cluster]

1
2
hadoop distcp hdfs://ip:9000/hive/spark-jars/metrics-core-4.1.1.jar /
hadoop distcp /spark/lib/HikariCP-2.5.1.jar hdfs://ip:9000/

-update -skipcrccheck

目录容量限制

hdfs dfs -count -q -h /tmp/a
目录个数,文件个数,文件总计大小

文件数量
hdfs dfsadmin -setQuota 2 /tmp/a
清除限制
hdfs dfsadmin -clrQuota /tmp/a

文件大小
hdfs dfsadmin -setSpaceQuota 5M /tmp/a
清除限制
hdfs dfsadmin -clrSpaceQuota /tmp/a

webhdfs使用

http://10.17.41.126:50070/explorer.html#/

user.name=hadoop

List

1
curl -i  "http://10.17.41.126:50070/webhdfs/v1/opendata?op=LISTSTATUS"

Create

1
curl -i -X PUT "http://10.17.41.126:50070/webhdfs/v1/opendata/testFile?op=CREATE&overwrite=false&replication=1"

返回datanode信息:
http://LGJF-ZYC6-HCY-SVR553.cmic:50075/webhdfs/v1/opendata/testFile?op=CREATE&namenoderpcaddress=LGJF-ZYC6-HCY-SVR553:9000&createflag=&createparent=true&overwrite=false&replication=1

1
curl -i -X PUT -T <LOCAL_FILE> "http://LGJF-ZYC6-HCY-SVR553.cmic:50075/webhdfs/v1/opendata/testFile?&op=CREATE&namenoderpcaddress=LGJF-ZYC6-HCY-SVR553:9000&createflag=&createparent=true&overwrite=false&replication=1"

脚本化可参考:https://github.com/pensz/hdfs_tools/tree/master/shell

控制容量 https://blog.csdn.net/weixin_30338481/article/details/94915052

机架感知

core-default.xml中增加配置

1
2
3
4
<property> 
<name>net.topology.script.file.name</name>
<value>/home/rack_topology.sh</value>
</property>

rack_topology.sh

1
2
3
4
5
6
7
8
9
10
11
12
TOPO=/mnt/d/topology.data

while [ $# -gt 0 ] ; do
nodeArg=$1
result=`awk -v var=${nodeArg} '$1 == var {print $2}' ${TOPO}`
shift
if [ -z "$result" ] ; then
echo -n "/default/rack "
else
echo -n "$result "
fi
done

topology.data

1
2
3
hadoopdata1.com     /dc1/rack1
hadoopdata1 /dc1/rack1
10.1.1.1 /dc1/rack2
1
2
3
4
5
6
7
8
import sys
rack = {dn178.tj":"rack1",
dn187.tj":"rack2"
"192.168.1.178":"rack1",
"192.168.1.187":"rack2",
}
if __name__=="__main__":
print "/" + rack.get(sys.argv[1],"rack0")

小文件合并

http://hadoop.apache.org/docs/current/hadoop-archives/HadoopArchives.html

短路读

https://hadoop.apache.org/docs/current/hadoop-project-dist/hadoop-hdfs/ShortCircuitLocalReads.html
https://hadoop.apache.org/docs/current/hadoop-project-dist/hadoop-common/NativeLibraries.html

外置客户端配置withKerberos

check: yarn app -list && hdfs dfs -ls .

core-site.xml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
  <configuration  xmlns:xi="http://www.w3.org/2001/XInclude">
<property>
<name>fs.defaultFS</name>
<value>hdfs://xx</value>
</property>
<!--kerberos认证环境-->
<property>
<name>hadoop.security.authentication</name>
<value>kerberos</value>
</property>
<!-- 用于支持同时访问kerberos和simple证环境 -->
<property>
<name>ipc.client.fallback-to-simple-auth-allowed</name>
<value>true</value>
</property>
</configuration>

hdfs-site.xml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
<configuration>

<property>
<name>dfs.nameservices</name>
<value>xx</value>
</property>

<property>
<name>dfs.ha.namenodes.xx</name>
<value>nn1,nn2</value>
</property>

<property>
<name>dfs.namenode.rpc-address.xx.nn1</name>
<value>h1:8020</value>
</property>

<property>
<name>dfs.namenode.rpc-address.xx.nn2</name>
<value>h2:8020</value>
</property>

<property>
<name>dfs.client.failover.proxy.provider.xx</name>
<value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
</property>

<property> --应该用不上
<name>dfs.namenode.kerberos.principal</name>
<value>nn/_HOST@HLWKDC</value>
</property>
</configuration>

yarn-site.xml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
<configuration>
<property>
<name>yarn.resourcemanager.ha.enabled</name>
<value>true</value>
</property>

<property>
<name>yarn.resourcemanager.ha.rm-ids</name>
<value>rm1,rm2</value>
</property>
<property>
<name>yarn.resourcemanager.hostname.rm1</name>
<value>h1-yy</value>
</property>

<property>
<name>yarn.resourcemanager.hostname.rm2</name>
<value>h2</value>
</property>
<property>
<name>yarn.client.failover-proxy-provider</name>
<value>org.apache.hadoop.yarn.client.ConfiguredRMFailoverProxyProvider</value>
</property>

<property>
<name>yarn.resourcemanager.principal</name>
<value>rm/_HOST@xx</value>
</property>
</configuration>

mapred-site.xml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
<configuration>
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
<property>
<name>mapreduce.job.queuename</name>
<value>root.xx</value>
</property>
<property>
<name>mapreduce.cluster.local.dir</name>
<value>/data/soft/hadoop/mrLocalTmp</value>
</property>
<property>
<name>mapreduce.application.classpath</name>
<value>$HADOOP_HOME/share/hadoop/mapreduce/*:$HADOOP_HOME/share/hadoop/mapreduce/lib/*</value>
</property>

<!-- local模式下不生效,限定为1,见源码org.apache.hadoop.mapred.LocalClientProtocolProvider-->
<property>
<name>mapreduce.local.map.tasks.maximum</name>
<value>10</value>
</property>
<property>
<name>mapreduce.local.reduce.tasks.maximum</name>
<value>10</value>
</property>
</configuration>

hive-site.xml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
<property>
<name>javax.jdo.option.ConnectionURL</name>
<value>jdbc:mysql://a:3306,b:3306,c:3306/hive?autoReconnect=true&amp;createDatabaseIfNotExist=true&amp;useUnicode=true&amp;characterEncoding=utf-8&amp;failOverReadOnly=false&amp;useSSL=false</value>
</property>
<property>
<name>javax.jdo.option.ConnectionDriverName</name>
<value>com.mysql.jdbc.Driver</value>
</property>
<property>
<name>javax.jdo.option.ConnectionUserName</name>
<value>xxx</value>
</property>
<property>
<name>javax.jdo.option.ConnectionPassword</name>
<value>xxx</value>
</property>

<property>
<name>hive.metastore.sasl.enabled</name>
<value>true</value>
</property>
<property>
<name>hive.metastore.kerberos.principal</name>
<value>hive/x@x</value>
</property>
<property>
<name>hive.metastore.uris</name>
<value>thrift://a:9083,thrift://b:9083,thrift://c:9083</value>
</property>
<property>
<name>hive.server2.authentication.kerberos.principal</name>
<value>hive/gzhlwtz@HLWKDC</value>
</property>

</configuration>
1
2
3
4
5
6
7
8
9
./start-thriftserver.sh \
--deploy-mode client \
--conf spark.driver.host=ip \
--name SparkJDBC \
--driver-memory 2g --executor-memory 4g --executor-cores 5 --num-executors 2 \
--hiveconf hive.server2.thrift.port=10001 \
--hiveconf hive.server2.authentication.kerberos.principal=x@x \
--hiveconf hive.server2.authentication.kerberos.keytab=/home/x.keytab \
--hiveconf hive.server2.enable.doAs=false

查看文件大小并排序

按统计大小排序
hdfs dfs -dus -h /hive/warehouse/orc_db/* |sed ‘s/ //‘| sort -h

按日期排序
hdfs dfs -ls -t /hive/warehouse/orc_db

修改副本数

(部分节点无法访问时 临时增加副本提供访问能力)

hdfs dfs -setrep 4 /hive/…
hadoop dfs -D dfs.replication=2 -put abc.txt /tmp

HAR

Apache Hadoop Archives – Hadoop Archives Guide

生成

1
2
3
4
hadoop archive -archiveName foo.har -p <src-parent-dir> [-r <replication factor>] <src>* <dest>

整个目录打包
hadoop archive -archiveName zoo.har -p /foo/bar -r 3 /outputdir

提交一个MapReduce任务来生成har文件
HAR文件实际上是一个以”.har”结尾命名的目录,其中至少包含三个文件:

  1. _index // 包内目录、文件的元数据信息
  2. _masterindex // 记录了“_index”文件
  3. part-00000 … // 直接拼接了原始文件内容,无压缩处理

读取

hdfs -ls har:///har/hp2.har/hp/

释放

串行 hdfs dfs -cp har:///user/zoo/foo.har/dir1 hdfs:/user/zoo/newdir
并行 hadoop distcp har:///user/zoo/foo.har/dir1 hdfs:/user/zoo/newdir

distcp

指定队列

hadoop distcp -Dmapred.job.queue.name=root.default …
hadoop distcp -Dmapreduce.job.queuename …

对拷sftp

sftpRemote=”/xxx”
sftp_user=”xxx”
sftp_pw=’xxx’ ##服务器密码
sftp_ip=””
sftp_port=””

hadoop distcp -D fs.sftp.impl=org.apache.hadoop.fs.sftp.SFTPFileSystem hdfs:///hive/warehouse/ads.db/user_label_iop/tp=202309 sftp://${sftp_user}:${sftp_pw}@${sftp_ip}:${sftp_port}${sftpRemote}/tpftp=202309

同时访问两个HA集群

core-site.xml
配置的当前客户端默认使用的nameservice

1
2
3
4
<property>
<name>fs.defaultFS</name>
<value>hdfs://nameservice1</value>
</property>

hdfs-site.xml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
<configuration>
<!-- services -->
<!-- local sevice and remote service -->
<property>
<name>dfs.nameservices</name>
<value>ns1,ns8</value>
</property>

<!-- remote namespace ns8 -->
<!-- service ns8 -->
<property>
<name>dfs.ha.namenodes.ns8</name>
<value>nn1,nn2</value>
</property>
<property>
<name>dfs.namenode.rpc-address.ns8.nn1</name>
<value>192.168.100.1:8020</value>
</property>
<property>
<name>dfs.namenode.rpc-address.ns8.nn2</name>
<value>192.168.100.2:8020</value>
</property>
<property>
<name>dfs.namenode.http-address.ns8.nn1</name>
<value>192.168.100.1:50070</value>
</property>
<property>
<name>dfs.namenode.http-address.ns8.nn2</name>
<value>192.168.100.2:50070</value>
</property>
<property>
<name>dfs.client.failover.proxy.provider.ns8</name>
<value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
</property>


<!-- local namespace ns1 -->
<!-- service ns1 -->
<property>
<name>dfs.internal.nameservices</name>
<value>ns1</value>
</property>
<property>
<name>dfs.ha.namenodes.ns1</name>
<value>nn1,nn2</value>
</property>
<property>
<name>dfs.namenode.rpc-address.ns1.nn1</name>
<value>dev01:8020</value>
</property>
<property>
<name>dfs.namenode.rpc-address.ns1.nn2</name>
<value>dev02:8020</value>
</property>
<property>
<name>dfs.namenode.http-address.ns1.nn1</name>
<value>dev01:50070</value>
</property>
<property>
<name>dfs.namenode.http-address.ns1.nn2</name>
<value>dev02:50070</value>
</property>
<property>
<name>dfs.client.failover.proxy.provider.ns1</name>
<value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
</property>
</configuration>
1
2
3
4
5
6
7
hadoop fs -ls /
hadoop fs -ls hdfs://ns1/ ##访问客户端默认的集群
hadoop fs -ls hdfs://ns8/ ##访问ns8集群

hadoop distcp -Dmapred.job.queue.name=root.userA -pb \
hdfs://ns1/user/userA/source_path/source_file \
hdfs://ns8/user/userA/dest_path

使用{}读取多个枚举值路径

${Burpoint_HDFS_PATH}/source={${data_sources}}/platform=*/year=$year/month=$month/day=$day

数据丢失处理

Safe mode is ON. The reported blocks 3 needs additional 2 blocks to reach the threshold 0.9990 of total blocks 5. The number of live datanodes 2 has reached the minimum number 0. Safe mode will be turned off automatically once the thresholds have been reached.

  • 确定zkfc启动 $HADOOP_HOME/sbin/hadoop-daemon.sh start zkfc
  • 退出安全模式 hdfs dfsadmin -safemode leave
  • 选定ActiveNN hdfs haadmin -failover nn1 nn2 (直接执行hdfs命令会报访问9000端口,standby状态不可读)
  • 检查问题文件 hdfs fsck /
  • 删除问题问题就 hdfs fsck /tmp -delete

balance

hdfs balancer -threshold 10 -blockpools BP-55455179-10.27.48.1-1677135619428 -source 10.27.48.20

threshold是控制每个DataNode的使用率不高于或者不低于集群平均的使用率

1
2
3
4
5
[-exclude [-f <hosts-file> | <comma-separated list of hosts>]]
[-include [-f <hosts-file> | <comma-separated list of hosts>]]
[-source [-f <hosts-file> | <comma-separated list of hosts>]]
[-blockpools <comma-separated list of blockpool ids>]
[-idleiterations <idleiterations>]
1
hdfs dfsadmin -setBalancerBandwidth <bandwidth in bytes per second>

200M: hdfs dfsadmin -setBalancerBandwidth 209715200

中间打印结果头为:
Time Stamp | Iteration# | Bytes Already Moved | Bytes Left To Move | Bytes Being Moved | NameNode

清空回收站

hdfs dfs -expunge

Web页面安全

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
  <property>
<name>hadoop.http.filter.initializers</name>
<value>org.apache.hadoop.security.AuthenticationFilterInitializer</value>
</property>
<property>
<name>hadoop.http.authentication.type</name>
<value>simple</value>
</property>
<property>
<name>hadoop.http.authentication.signature.secret.file</name>
<value>/opt/hadoop/secret/hadoop-http-auth-signature-secret</value>
</property>
<property>
<name>hadoop.http.authentication.simple.anonymous.allowed</name>
<value>false</value>
</property>

访问地址为secret中写入的字符 http://127.0.0.1:9870?user.name=qazwsx$123

通过nginx配置用户名密码

yum install httpd-tools
yum install httpd
启动 httpd 命令:service httpd restart
httpd 的默认安装目录在:/etc/httpd/
关于其配置可以自行查看 /etc/httpd/httpd.conf 文件
如果启动成功后,访问服务器的80端口会出现apache的welcome界面。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
# 设置用户名,密码 生成 db文件
htpasswd -c /usr/local/nginx/passwd.db username password
# 查看生成的db文件内容
cat /usr/nginx/conf/htpasswd.users

vi /usr/local/nginx/conf/nginx.conf
server {
listen 50070;
server_name localhost;

location / {
auth_basic "hadoop001"; # 虚拟主机认证命名
auth_basic_user_file /usr/local/nginx/passwd.db; # 虚拟主机用户名密码认证数据库
proxy_pass http://127.0.0.1:9870; # hadoop 访问
}
}

数据权限限制

https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-hdfs/HdfsPermissionsGuide.html

  • NameNode启动进程者默认为superUser,不受限制 (禁止其他用户启动集群)
  • 普通用户使用自身账号访问文件,开放hadoop/spark/flink命令、hdfs目录、个人代码空间权限
  • 对期望限制的文件修改权限,后续文件管理由文件权限所有者或superUser处理
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
dfs.permissions.enabled = true -> hdfs-site 默认值true

dfs.permissions.superusergroup = supergroup -> hdfs-site 默认值supergroup

fs.permissions.umask-mode = 022 -> core-site 默认值022

# hadoop-policy.xml
yarn rmadmin -refreshServiceAcl
hadoop dfsadmin -refreshServiceAcl

基础开关
hadoop.security.authorization=true
控制提交任务/ JobTracker
security.job.client.protocol.acl
访问HDFS /NameNode
security.client.protocol.acl


# yarn-site

开启yarn ACLs:
hadoop: core-site.xml
hadoop.security.authorization=true #开启服务级别验证,否则hadoop组件的acl设置不生效
yarn: yarn-site.xml
<property>
<name>yarn.acl.enable</name>
<value>true</value>
</property>
<property>
<name>yarn.admin.acl</name>
<value>hdfs</value>
</property>

<property>
<name>yarn.nodemanager.default-container-executor.log-dirs.permissions</name>
<value>755</value>
</property>
$ vi $HADOOP_CONF_DIR/capacity-scheduler.xml

$ yarn rmadmin -refreshQueues
yarn.scheduler.capacity.root.<queue-path>.acl_submit_applications

使用者操作方式

启动:使用hdfs用户启动hadoop服务

任务部署:ds, 上传jar包、运行任务都将以hadoop运行,生产是数据hadoop写出为755

个人调试/查看数据:可以本地执行hadoop/spark/flink命令执行

数据保护:定期修改关键数据为hdfs用户下755,只有hdfs可删除文件

hdfs dfs -chown hdfs xxx

hdfs dfs -chmod  -R 755 xxx

配置权限

hdfs dfs -getfacl /

hdfs dfs -setfacl -m user:hue:rwx /warehouse/tablespace/managed/hive

GroupMapping配置

参考Hadoop官网配置分组,如果没有配置就fallback到User级别

core-site.xml优先按照配置的静态映射确定组,默认值为dr.who=;

1
2
3
4
<property>
<name>hadoop.user.group.static.mapping.overrides</name>
<value>user1=group1,group2;user2=;user3=group2</value>
</property>

无匹配的静态项则使用配置映射service provider为hadoop.security.group.mapping

配置静态映射后刷新配置:
hdfs dfsadmin -refreshUserToGroupsMappings

–只配置nn,但需要重启nn

hadoop用户和权限 - 过雁 - 博客园 (cnblogs.com)

HDFS文件清理巡检

防止使用者乱放文件

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
# modified 20230303
KNOWN_FILE="^/data$"
KNOWN_FILE=$KNOWN_FILE"|^/dolphinscheduler$"
KNOWN_FILE=$KNOWN_FILE"|^/flink$"
KNOWN_FILE=$KNOWN_FILE"|^/hive$"
KNOWN_FILE=$KNOWN_FILE"|^/spark$"
KNOWN_FILE=$KNOWN_FILE"|^/tmp$"
KNOWN_FILE=$KNOWN_FILE"|^/user$"
KNOWN_FILE=$KNOWN_FILE"|^/yarn$"
hdfs dfs -ls -C / |grep -Ev $KNOWN_FILE

KNOWN_FILE="^/data/huadan$"
KNOWN_FILE=$KNOWN_FILE"|^/data/env$"
hdfs dfs -ls -C /data |grep -Ev $KNOWN_FILE

KNOWN_FILE="^/dolphinscheduler/hadoop$"
hdfs dfs -ls -C /dolphinscheduler |grep -Ev $KNOWN_FILE

KNOWN_FILE="^/flink/completed-jobs$"
hdfs dfs -ls -C /flink |grep -Ev $KNOWN_FILE

KNOWN_FILE="^/spark/warehouse$"
KNOWN_FILE=$KNOWN_FILE"|^/spark/spark-history$"
KNOWN_FILE=$KNOWN_FILE"|^/spark/spark-jar.zip$"
hdfs dfs -ls -C /spark |grep -Ev $KNOWN_FILE

KNOWN_FILE="^/hive/warehouse$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/tmp$"
hdfs dfs -ls -C /hive |grep -Ev $KNOWN_FILE


KNOWN_FILE="^/hive/warehouse/ads.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/default.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/dim.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/dm.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/dw.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/dwd.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/lg.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/mid.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/ods.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/oracle_orc_fromdatax.db$"
KNOWN_FILE=$KNOWN_FILE"|^/hive/warehouse/tmp.db$"
hdfs dfs -ls -C /hive/warehouse |grep -Ev $KNOWN_FILE


查询在读写的hive表

1
grep '^2024-01-10'  /data02/hadoop/log/hadoop-hadoop-namenode-master.log|grep completeFile  |awk '{print $7}' |grep '^/hive/warehouse' |awk -F'/' '{print $1"/"$2"/"$3"/"$4"/"$5}'|sort -u

文件使用检测-审计

背景: 大量数据文件生成,哪些是不用的,需要判断以清理或冷数据处理

配置debug级别才打印的命令(查询表的时候是 cmd=open)

1
dfs.namenode.audit.log.debug.cmdlist=getfileinfo,listStatus

开启审计(需要重启NN,因为会用在jvm启动命令上)

1
export HDFS_AUDIT_LOGGER=INFO,RFAAUDIT

$HADOOP_HOME/sbin/hadoop-daemon.sh stop namenode
$HADOOP_HOME/sbin/hadoop-daemon.sh start namenode

所有访问IP

awk ‘{print $8}’ hdfs-audit.log | sort -u

远程拉取的文件是否在用

所有create操作
cat hdfs-audit.log |grep cmd=create |awk ‘{print $8,$9,$10}’ | awk -F’/‘ ‘{print $2”/“$3”/“$4”/“$5”/“$6}’ |sort -u > check_create

所有open操作
cat hdfs-audit.log |grep cmd=open |grep -v application | grep -v inprogress |awk ‘{print $8,$9,$10}’ | awk -F’/‘ ‘{print $2”/“$3”/“$4”/“$5”/“$6}’ |sort -u > check_open

查看远程集群create的表 /hive/warehouse/xx.db/xxx
grep “10.27.48” check_create |awk -F’/‘ ‘{print “/“$2”/“$3”/“$4”/“$5}’ |sort -u

查看本地集群open的表 /hive/warehouse/xx.db/xxx
grep -v “10.27.48” check_open |grep -v Staging | awk -F’/‘ ‘{print “/“$2”/“$3”/“$4”/“$5}’ |sort -u

校验
grep KEYWORDS hdfs-audit.log |awk ‘{print $8,$9,$10}’

create的文件是否在用

查看当前打开的文件
hdfs dfsadmin -listOpenFiles

表名前缀分组统计

1
hdfs dfs -ls /hive/warehouse/tmp.db/ |awk '{print $8}' |awk -F'/' '{print $5}' |awk -F'_' '{print $1}' | sort |uniq -c |sort -h

节点管理-退役

Apache Hadoop 3.3.6 – HDFS DataNode Admin Guide

使用场景:节点退服、节点维护,作用是告诉namenode该节点后续的安排以控制数据副本复制的操作,比如维护状态就明确节点只是临时下线不需要进行副本复制,减少不必要的IO

配置方式1: Hostname-only

该方式只支持decommission and recommission,不支持maintenance

echo “10.17.41.133” > datanode.excludes
echo “10.17.41.133” > datanode.includes

vi hdfs-site.xml (dfs.hosts and dfs.hosts.exclude)

1
2
3
4
<property> 
<name>dfs.hosts.exclude</name>
<value>/data/hadoop-2.8.3/etc/hadoop/datanode.excludes</value>
</property>

hdfs dfsadmin -refreshNodes

显示过滤后的16行
hdfs dfsadmin -report |grep -A 16 10.17.41.26

配置方式2: Json

hdfs-site.xml修改以下classname需要重启NN

hdfs-site.xml
1
2
3
4
5
6
7
8
<property> 
<name>dfs.namenode.hosts.provider.classname</name>
<value>org.apache.hadoop.hdfs.server.blockmanagement.CombinedHostFileManager</value>
</property>
<property>
<name>dfs.hosts</name>
<value>/data/soft/hadoop/etc/hadoop/datanode.json</value>
</property>

不配置Normal状态的节点,表示所有其他节点都可以注册到Namenode
admin state. The default value is NORMALDECOMMISSIONED for decommission; IN_MAINTENANCE for maintenance state.

datanode.json
1
2
3
4
5
6
7
[
{
"hostName": "10.27.48.2",
"maintenanceExpireTimeInMS": "120000", # The default value is forever.
"adminState": "IN_MAINTENANCE"
}
]
1
2
3
4
5
6
7
8

$HADOOP_HOME/sbin/hadoop-daemon.sh stop namenode
$HADOOP_HOME/sbin/hadoop-daemon.sh start namenode
$HADOOP_HOME/sbin/hadoop-daemon.sh stop datanode
$HADOOP_HOME/sbin/hadoop-daemon.sh start datanode
$HADOOP_HOME/sbin/hadoop-daemon.sh stop nodemanager
$HADOOP_HOME/sbin/hadoop-daemon.sh start nodemanager

主备切换

hdfs haadmin -getAllServiceState
hdfs haadmin -failover nn1 nn2

无须重启 刷新配置

reconfigureProperty比如fs.protected.directories指定保护以免删除的目录

hdfs dfsadmin -reconfig <datanodenamenode <host:ipc_port> <start|status|properties>]

properties命令查看所有支持刷新的配置项,start启动刷新,status查看结果