kerberos - GSS 异常 : No valid credentials provided (Mechanism level: Failed to find any Kerberos tgt) while connecting Polybase with Kerberos

Question

我们想通过 Polybase 将我们的 SQL Server 2016 Enterprise 与我们的 Kerberized OnPrem Hadoop-Cluster 和 Cloudera 5.14 连接起来。

我按照 Microsoft PolyBase Guide 配置 Polybase。在这个主题上工作了几天后，由于出现异常，我无法继续:javax.security.sasl.SaslException: GSS initiate failed [Caused by GSSException: No valid credentials provided (Mechanism level: Failed to find any Kerberos tgt)]

Microsoft 有一个内置的 troubleshooting the connectivity with PolyBase and Kerberos 诊断工具。在 Microsoft 的这份故障排除指南中，有 4 个检查点，我被困在检查点 4 上。有关检查点(我成功的地方)的简短信息:

• 检查点 1:成功! 针对 KDC 进行身份验证并收到 TGT
• 检查点 2:成功!关于故障排除指南，PolyBase 将尝试访问 HDFS 但失败，因为请求不包含必要的服务票证。
• 检查点 3:成功!第二个十六进制转储表明 SQL Server 成功使用了 TGT 并从 KDC 获取了适用于名称节点 SPN 的服务票证。
• 检查点 4:未成功 Hadoop 使用 ST(服务票证)对 SQL Server 进行了身份验证，并授予 session 访问安全资源的权限。

krb5.conf文件

[libdefaults]
default_realm = COMPANY.REALM.COM
dns_lookup_kdc = false
dns_lookup_realm = false
ticket_lifetime = 86400
renew_lifetime = 604800
forwardable = true
default_tgs_enctypes = aes256-cts-hmac-sha1-96 aes128-cts-hmac-sha1-96
default_tkt_enctypes = aes256-cts-hmac-sha1-96 aes128-cts-hmac-sha1-96
permitted_enctypes = aes256-cts-hmac-sha1-96 aes128-cts-hmac-sha1-96
udp_preference_limit = 1
kdc_timeout = 3000
[realms]
COMPANY.REALM.COM = {
kdc = ipadress.kdc.host
admin_server = ipadress.kdc.host
}
[logging]
default = FILE:/var/log/krb5/kdc.log
kdc = FILE:/var/log/krb5/kdc.log
admin_server = FILE:/var/log/krb5/kadmind.log

SQL-Server 上 Polybase 的 core-site.xml

<?xml version="1.0" encoding="utf-8"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<!-- Put site-specific property overrides in this file. -->
<configuration>
  <property>
    <name>io.file.buffer.size</name>
    <value>131072</value>
  </property>
  <property>
    <name>ipc.client.connect.max.retries</name>
    <value>2</value>
  </property>
  <property>
    <name>ipc.client.connect.max.retries.on.timeouts</name>
    <value>2</value>
  </property>

<!-- kerberos security information, PLEASE FILL THESE IN ACCORDING TO HADOOP CLUSTER CONFIG -->
<property>
    <name>polybase.kerberos.realm</name>
    <value>COMPANY.REALM.COM</value>
  </property>
  <property>
    <name>polybase.kerberos.kdchost</name>
    <value>ipadress.kdc.host</value>
  </property>
  <property>
    <name>hadoop.security.authentication</name>
    <value>KERBEROS</value>
  </property>
</configuration>

用于 SQL-Server 上的 Polybase 的 hdfs-site.xml

<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<!-- Put site-specific property overrides in this file. -->
<configuration>
  <property>
    <name>dfs.block.size</name>
    <value>268435456</value> 
  </property>
  <!-- Client side file system caching is disabled below for credential refresh and 
       settting the below cache disabled options to true might result in 
       stale credentials when an alter credential or alter datasource is performed
  -->
  <property>
    <name>fs.wasb.impl.disable.cache</name>
    <value>true</value>
  </property>
  <property>
    <name>fs.wasbs.impl.disable.cache</name>
    <value>true</value>
  </property>
  <property>
    <name>fs.asv.impl.disable.cache</name>
    <value>true</value>
  </property>
  <property>
    <name>fs.asvs.impl.disable.cache</name>
    <value>true</value>
  </property>
  <property>
    <name>fs.hdfs.impl.disable.cache</name>
    <value>true</value>
  </property>
<!-- kerberos security information, PLEASE FILL THESE IN ACCORDING TO HADOOP CLUSTER CONFIG -->
  <property>
    <name>dfs.namenode.kerberos.principal</name>
    <value>hdfs/_HOST@COMPANY.REALM.COM</value> 
  </property>
</configuration>

多碱基异常

[2018-06-22 12:51:50,349] WARN  2872[main] - org.apache.hadoop.security.UserGroupInformation.hasSufficientTimeElapsed(UserGroupInformation.java:1156) - Not attempting to re-login since the last re-login was attempted less than 600 seconds before.
[2018-06-22 12:51:53,568] WARN  6091[main] - org.apache.hadoop.security.UserGroupInformation.hasSufficientTimeElapsed(UserGroupInformation.java:1156) - Not attempting to re-login since the last re-login was attempted less than 600 seconds before.
[2018-06-22 12:51:56,127] WARN  8650[main] - org.apache.hadoop.security.UserGroupInformation.hasSufficientTimeElapsed(UserGroupInformation.java:1156) - Not attempting to re-login since the last re-login was attempted less than 600 seconds before.
[2018-06-22 12:51:58,998] WARN 11521[main] - org.apache.hadoop.security.UserGroupInformation.hasSufficientTimeElapsed(UserGroupInformation.java:1156) - Not attempting to re-login since the last re-login was attempted less than 600 seconds before.
[2018-06-22 12:51:59,139] WARN 11662[main] - org.apache.hadoop.ipc.Client$Connection$1.run(Client.java:676) - Couldn't setup connection for hdfs@COMPANY.REALM.COM to IPADRESS_OF_NAMENODE:8020
javax.security.sasl.SaslException: GSS initiate failed [Caused by GSSException: No valid credentials provided (Mechanism level: Failed to find any Kerberos tgt)]

NameNode 上的日志条目

Socket Reader #1 for port 8020: readAndProcess from client IP-ADRESS_SQL-SERVER threw exception [javax.security.sasl.SaslException: GSS initiate failed [Caused by GSSException: Failure unspecified at GSS-API level (Mechanism level: AES128 CTS mode with HMAC SHA1-96 encryption type not in permitted_enctypes list)]]

Auth failed for IP-ADRESS_SQL-SERVER:60484:null (GSS initiate failed) with true cause: (GSS initiate failed)

让我感到困惑的部分是来 self 们的 NameNode 的日志条目，因为 AES128 CTS 模式与 HMAC SHA1-96 已经在允许的 enctypes 列表中，如 krb5.conf 和 Cloudera Manager UI 中所示

感谢您的帮助!

Answer 1

问题在我们重新启动集群后自行解决。我认为问题是由于某些正在运行的服务，我们的 Hadoop-Cluster 中的 krb5.conf 文件无法分布在所有节点上。 Cloudera Manager 中也有关于 Kerberos 的陈旧配置的警告。非常感谢大家!