1

我确实对 Kerberos 凭据有疑问。这项工作基于集群，并且在每个数据节点上都提供了密钥表。基本上它是一个 oozie 工作流 shell 操作，其目的是通过 spark 作业写入 HBase。如果作业在没有 oozie 的集群模式下运行，它会按预期工作。但是对于 oozie，它会抛出一个异常，如下所示：

WARN AbstractRpcClient: Exception encountered while connecting to the server 
: javax.security.sasl.SaslException: GSS initiate failed [Caused by 
GSSException: No valid credentials provided (Mechanism level: Failed to find 
any Kerberos tgt)]
18/11/26 15:30:24 ERROR AbstractRpcClient: SASL authentication failed. The 
most likely cause is missing or invalid credentials. Consider 'kinit'.
javax.security.sasl.SaslException: GSS initiate failed [Caused by 
GSSException: No valid credentials provided (Mechanism level: Failed to find 
any Kerberos tgt)]
at 
com.sun.security.sasl.gsskerb.GssKrb5Client.evaluateChallenge
(GssKrb5Client.java:211)
at 
org.apache.hadoop.hbase.security.HBaseSaslRpcClient.
saslConnect(HBaseSaslRpcClient.java:179)
at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection
.setupSaslConnection(RpcClientImpl.java:611)
at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection
.access$600(RpcClientImpl.java:156)
at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection$2
.run(RpcClientImpl.java:737)
at org.apache.hadoop.hbase.ipc.RpcClientImpl$Connection$2
.run(RpcClientImpl.java:734)
at java.security.AccessController.doPrivileged(Native Method)

oozie shell 动作如下所示：

<action name="spark-hbase" retry-max="${retryMax}" retry-interval="${retryInterval}">
<shell xmlns="uri:oozie:shell-action:0.3">
  <exec>submit.sh</exec>
  <env-var>QUEUE_NAME=${queueName}</env-var>
  <env-var>PRINCIPAL=${principal}</env-var>
  <env-var>KEYTAB=${keytab}</env-var>
  <env-var>VERBOSE=${verbose}</env-var>
  <env-var>CURR_DATE=${firstNotNull(currentDate, "")}</env-var>
  <env-var>DATA_TABLE=${dataTable}</env-var>
  <file>bin/submit.sh</file>
</shell>
<ok to="end"/>
<error to="kill"/>
</action>

submit.sh 文件的 spark-submit 命令如下所示：

enter code here
CLASS="App class location"
JAR="compiled jar file"

HBASE_JARS="HBase jars"
HBASE_CONF='hbase-site.xml location'

HIVE_JARS="Hive jars"
HIVE_CONF='tez-site.xml location'

HADOOP_CONF='hdfs-site.xml location'

SPARK_BIN_DIR="spark2-client bin directory location"

${SPARK_BIN_DIR}/spark-submit \
  --class ${CLASS} \
  --principal "${PRINCIPAL}" \
  --keytab "${KEYTAB}" \
  --master yarn \
  --deploy-mode cluster \
  --driver-memory 10G \
  --executor-memory 4G \
  --num-executors 10 \
  --conf spark.default.parallelism=24 \
  --jars ${HBASE_JARS},${HIVE_JARS} \
  --files ${HBASE_CONF},${HIVE_CONF},${HADOOP_CONF} \
  --conf spark.ui.port=4042 \
  --conf "spark.executor.extraJavaOptions=-verbose:class - 
  Dsun.security.krb5.debug=true" \
  --conf "spark.driver.extraJavaOptions=-verbose:class - 
  Dsun.security.krb5.debug=true" \
  --queue "${QUEUE_NAME}" \
  ${JAR} \
    --app.name "spark-hbase" \
    --data.table "${DATA_TABLE}" \
    --verbose
4

1 回答 1

0

在集群中的所有节点上创建软链接可能并不总是可行的。我们通过在 spark-submit 命令之前覆盖 shell 中的 SPARK_CONF_DIR 环境变量来在 spark 配置中添加 hbase 配置目录来解决它。

export SPARK_CONF_DIR=/etc/spark2/conf:/etc/hbase/conf
于 2019-02-23T11:04:15.983 回答