https://zeppelin.apache.org/download.html

cp conf/zeppelin-site.xml.template conf/zeppelin-site.xml
vi conf/zeppelin-site.xml
修改zeppelin-site.xml指定绑定的ip和port

bin/zeppelin-daemon.sh start

bin/zeppelin-daemon.sh stop

支持的组件

alluxio    elasticsearch  groovy        java     kylin    neo4j   scalding      sparql
angular    file           hazelcastjet  jdbc     lens     pig     scio          submarine
beam       flink          hbase         jupyter  livy     python  sh            zeppelin-interpreter-shaded-0.10.1.jar
bigquery   flink-cmd      ignite        kotlin   md       r       spark
cassandra  geode          influxdb      ksql     mongodb  sap     spark-submit

解释器使用credentials密码加密

配置文件添加配置项 zeppelin.credentials.encryptKey,注意密钥需要是16/24/32字节
创建interpreter时不配置默认密码，用户添加credentials即可

SparkThriftServer方案

配置个人账密连接受限用户(proxyuser)启动的thrift，thrift可显示sql对应用户，文件操作为受限用户

配置JDBC的依赖为 org.apache.hive:hive-jdbc:2.3.4 会自动联网下载. (离线部署则使用hive-jdbc-2.3.10-standalone.jar，否则还需要下载其他依赖)
driver为 org.apache.hive.jdbc.HiveDriver

./local-repo/org/apache/hive/hive-jdbc
./local-repo/org/apache/hive/hive-jdbc/2.3.4/hive-jdbc-2.3.4.pom
./local-repo/org/apache/hive/hive-jdbc/2.3.4/hive-jdbc-2.3.4.pom.sha1
./local-repo/org/apache/hive/hive-jdbc/2.3.4/hive-jdbc-2.3.4.jar
./local-repo/org/apache/hive/hive-jdbc/2.3.4/hive-jdbc-2.3.4.jar.sha1
./local-repo/query-48/hive-jdbc-2.3.4.jar
./local-repo/query-194/hive-jdbc-2.3.4.jar

使用登录用户身份连接，但还是用的STS进程用户查询，添加自定义配置项 default.proxy.user.property=hive.server2.proxy.user
hive.server2.enable.doAs=true

为了能避免创建目录是当前登录用户、写出数据又是启动STS的用户导致permission deny问题，配置hive.server2.enable.doAs=false ，使得操作都是sts的启动用户

最后使用的方案是启动thriftserver时指定--proxy-user zeppelin，则都用zeppelin身份进行io，而在thriftserver可以知道谁执行

zeppelin.spark.run.asLoginUser

global share: 以第一个使用的人的身份启动
isolated per note: 每个人都独立启动
scoped per note: 同一个sparksession，但变量相互不干扰

检查用户登陆

cat zeppelin-hadoop-LGJF-ZYC6-HCY-SVR556.log2023-08-0* |grep user |awk -F ‘:’ ‘{print $8}’ |sort -u

grep “Creating connection pool” zeppelin-interpreter-query-shared_process-hadoop-a51-dg-hcy-bi-cal-003.log |awk -F’,’ ‘{print $3}’

用户密码加密

ini方式

shiro.ini


[main]
matcher=org.apache.shiro.authc.credential.HashedCredentialsMatcher
matcher.hashAlgorithmName=sha-256
matcher.storedCredentialsHexEncoded=true
## 加密次数
#matcher.hashIterations=2
## 存储散列后的密码是否为16进制  # true = hex, false = base64


myRealm.credentialsMatcher = $matcher

[users] 
# user1 = sha256-hashed-hex-encoded password, role1, role2, ... 
user1 = 2bb80d537b1da3e38bd30361aa855686bde0eacd7162fef6a25fe97bf527a25b, role1, role2, ...

生成密码方法： echo -n thisIsPassword | sha256sum

其他验证类：
SimpleCredentialsMatcher
AllowAllCredentialsMatcher
HashedCredentialsMatcher
PasswordMatcher

JDBC方式

Maven Repository: org.xerial » sqlite-jdbc

[main] 
myRealm=org.apache.shiro.realm.jdbc.JdbcRealm
#明文就 SimpleCredentialsMatcher Sha256CredentialsMatcher
myRealmCredentialsMatcher=org.apache.shiro.authc.credential.SimpleCredentialsMatcher
myRealm.credentialsMatcher=$myRealmCredentialsMatcher

#配置数据库的信息
dataSource=org.sqlite.SQLiteDataSource
dataSource.url=jdbc:sqlite:/data/zeppelin/user.db
myRealm.dataSource=$dataSource

sqlite3 user.db
create table users (username text, password text);
insert into users values ('xx','xx'); -- sha256+hex

create table user_roles(username text, role_name text);
insert into user_roles values ('xx','admin');
update user_roles set role_name='rol' where username='xx';

create table roles_permissions (username text, permission text);

代码

ZeppelinServer

NotebookService.java[runParagraph] 界面发起执行
Note.run
Paragraph.execute getBindedInterpreter()
InterpreterFactory#getInterpreter
InterpreterSetting getDefaultInterpreter(ExecutionContext executionContext)
InterpreterSetting#getOrCreateSession(org.apache.zeppelin.interpreter.ExecutionContext)
InterpreterSetting.java[getOrCreateInterpreterGroup] Create InterpreterGroup with groupId
ManagedInterpreterGroup#getOrCreateSession
InterpreterSetting#createInterpreters : new RemoteInterpreter
Paragraph.execute interpreter.getScheduler().submit(this);放入队列

RemotetScheduler(每个interpreter一个) /AbstractScheduler#run出队列->runJobInScheduler->JobRunner->runJob
Paragraph/InterpretJob(是一种job) jobRun
RemoteInterpreter.interpret发起执行

getOrCreateInterpreterProcess
ManagedInterpreterGroup.java[getOrCreateInterpreterProcess]
- interpreterSetting.createInterpreterProcess -> createLauncher
- PluginManager.java[loadInterpreterLauncher]
- StandardInterpreterLauncher.java[launchDirectly]
- ProcessLauncher.java[transition] 启动interpreter进程
- RemoteInterpreterProcess#callRemoteFunction 发起远程执行
RemoteInterpreterServer 作为进程容器，后续由zeppelinServer的RemoteInterpreterEventClient作为客户端创建Interpreter
- JDBCInterpreter 其中一种interpreter实例

zeppelin