1.1 Common Faults
1.1.1 Authentication Fails When You Run the Oozie Command
Applicable Versions
V100R002C50SPC20*,
V100R002C60SPC20*,
V100R002C60U10,
V100R002C60U10SPC00*,
V100R002C60U20,
V100R002C60U20SPC00*,
V100R002C70SPC20*
Diagnostic
When you run the oozie admin -oozie https://xx.xx.xx.xx:21003/oozie/ -run command, the following information is displayed or a message indicating that the authentication fails is reported. No permission error exists:
"Error: AUTHENTICATION : Could not authenticate, GSSException: No valid credentials provided (Mechanism level: Server not found in Kerberos database (7) - UNKNOWN_SERVER)"
Fault Locating
1. The Kerberos authentication is not performed before you submit a job using the Oozie client in the security version.
2. The user who submits the Oozie job does not have the permission.
Solution
Step 1
Run the kinit command to authenticate the user name and then submit the job.
Step 2
Add a common user role of Oozie, queue the submission permission of Yarn and HDFS access permission to the user.
1.1.2 403 Error Is Reported When Oozie Submits a Job
Applicable Versions
V100R002C50SPC20*,
V100R002C60SPC20*,
V100R002C60U10,
V100R002C60U10SPC00*,
V100R002C60U20,
V100R002C60U20SPC00*,
V100R002C70SPC20*
Symptom
Oozie fails to submit a job and the 403 error is reported. The submission methods are CLI and Java API.
Fault Locating
The cause of this issue has been described in the product documentation. That is, Oozie and Solr are installed on the same node and the job is submitted by using the IP address of the active Oozie rather than the host name.
Solution
Step 1
Submit the job by using the host name rather than the IP address of the host where Oozie resides.
Oozie job -oozie https://Oozie active node hostname:21003/oozie -config job.properties -run
1.1.3 One Node Is Successfully Started, and One Node Fails to Be Started
Applicable Versions
V100R002C50SPC20*,
V100R002C60SPC20*,
V100R002C60U10,
V100R002C60U10SPC00*,
V100R002C60U20,
V100R002C60U20SPC00*,
V100R002C70SPC20*
Diagnostic
One of the Oozie nodes is successfully started and the other one fails to be started. When the failed instance is switched over to the first node, the start succeeds.
Fault Locating
Some default Tomcat environment variables, such as CATALINA_BASE and CATALINA_HOME, are manually modified. As a result, paths fail to be loaded during the Oozie start process.
Solution
Step 1
Run the unset CATALINA_BASE command.
Step 2
Modify the definition of the environment variable CATALINA_BASE in /etc/profile.
Step 3
Restart the NodeAgent.
su - omm
/opt/huawei/Bigdata/nodeagent/bin/stop-agent.sh;/opt/huawei/Bigdata/nodeagent/bin/start-agent.sh
1.1.4 Oozie Fails to Submit a Loader Job
Applicable Versions
V100R002C50SPC20*,
V100R002C60SPC20*,
V100R002C60U10,
V100R002C60U10SPC00*,
V100R002C60U20,
V100R002C60U20SPC00*,
V100R002C70SPC20*
Diagnostic
The following error message is reported when Oozie fails to submit a Loader job:
Failing Oozie Launcher, Main class [org.apache.oozie.action.hadoop.SqoopMain], main() threw exception, Server has returned exception
org.apache.sqoop.common.SqoopException: Server has returned exception
sqoop.float.ip:null
sqoop.https.port:null
Fault Locating
1. The Oozie user who submits the Loader job does not have the admin rights.
2. During cluster installation, Oozie is installed before Loader. When the configuration file of Oozie is generated, necessary information, such as the floating IP address and port of Loader, is not contained in the Oozie configuration item. As a result, these necessary parameters are left blank during job submission, causing submission failures.
Solution
Step 1
Add the admin rights of Loader to the Oozie user who submits the job.
Figure 1-1 Adding a role
Step 2
Restart the Oozie component to generate a new configuration file.
