This is the documentation for Cloudera Manager 4.8.3.
Documentation for other versions is available at Cloudera Documentation.

Known Issues Fixed in Cloudera Manager 4.0.1

— (Applies to CDH4 only) When HDFS HA is enabled but Automatic Failover is disabled, both NameNodes may be left in standby mode upon restart.

If you have HDFS High Availability enabled but you do not have Automatic Failover enabled, the HDFS service will have both NameNodes in standby mode when you do any of the following: stop or restart the HDFS service, restart the active NameNode, or stop and start the entire cluster.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: Use the Manual Failover command on the Action menu for the HDFS service to activate one of the NameNodes.

— (Applies to CDH4 only) If HDFS High Availability Automatic Failover is enabled, manual failover is not possible.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: None

— Using yum to upgrade a managed node that is not running the Cloudera Manager server will erroneously install the cloudera-manager-server package.

If you use yum update to upgrade a node that is not running the Cloudera Manager server (instead of using the upgrade wizard in Cloudera Manager) it installs the cloudera-manager-server package, even though it is not needed.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: Either remove the server package immediately after the upgrade, or explicitly upgrade 'cloudera-manager-daemons' and 'cloudera-manager-agent' instead of upgrading everything.

— (Applies to CDH4 only) With HDFS High Availability or Federation enabled, the File Browser may encounter problems.

There are known issues if the Hue File Browser is used with High Availability or Federation enabled, because the File Browser needs to use HttpFS but it cannot.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: To use the Hue File Browser with High Availability enabled, point Hue at the active NameNode manually. If a failover occurs, the Hue File Browser may fail. To use the Hue File Browser with Federation enabled, you can only point Hue at individual NameNodes, not the entire Federated namespace.

— (Applies to CDH4 only) If HDFS Federation is enabled, it is not possible to create a directory directly off the root directory in an individual namespace.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: Point the 'hdfs' client application directly at the individual namespace, rather than at the entire federation viewfs.

— The HBase Master Web UI link does not work for standby masters.

As of Cloudera Manager 3.7.3, you can configure multiple HBase masters. In the Role Instance Status page for running HBase masters, please note that the HBase Web UI link only works for the active master of the HBase service. Standby (backup) HBase masters do not expose a web interface, so the link does not work.

Severity: Low

Anticipated resolution: Fixed in Cloudera Manager 4.0.1

Workaround: None

— ZooKeeper server does not remove snapshot and log files.

The ZooKeeper server does not remove the snapshots and log files, so they will accumulate over time. You will need to cleanup this directory occasionally, based on your backup schedules and processes.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1. You can manually remove the snapshots and log files by using the Cleanup command on the Actions menu of the ZooKeeper service.

Workaround: To automate the cleanup, a zkCleanup.sh script is provided in the bin directory of the hadoop-zookeeper base package. Modify this script as necessary for your situation. In general, you want to run this as a cron task based on your backup schedule. For more information, see ZooKeeper Installation documentation in CDH3 Installation Guide.

— Cloudera Manager does not provide options for configuring Beeswax to use a Hive Metastore backed by MySQL.

Cloudera Manager currently does not support configuration options to allow Beeswax to use a MySQL Hive metastore. However, you can use the Safety Valve for the Beeswax Server to add these configuration options.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: Add options to the Beeswax Server Hive Configuration Safety Valve:

In the Cloudera Manager Admin Console, navigate to your Hue Service instance, then to the Beeswax Server instance and click the Configuration tab to show the configuration options.

Find the Beeswax Server Hive Configuration Safety Valve and click on the (blank) edit field.

Add the following lines (where myhivehost and myhivepassword are replaced by your own hostname and password):

                  <!-- Hive Execution Parameters --> <property>
                  <name>javax.jdo.option.ConnectionURL</name>
                  <value>jdbc:mysql://myhivehost/hive</value> </property> <property>
                  <name>javax.jdo.option.ConnectionDriverName</name>
                  <value>com.mysql.jdbc.Driver</value> </property> <property>
                  <name>javax.jdo.option.ConnectionUserName</name> <value>hive</value>
                  </property> <property> <name>javax.jdo.option.ConnectionPassword</name>
                  <value>myhivepassword</value> </property>
                

Click Save Changes

Restart the entire Hue service.

— The default heap size for the Cloudera Manager Server is now 2GB.

The default heap size for the Cloudera Manager Server has been increased from 1GB to 2GB.

You can configure the maximum heap size of the SCM process by editing the CMF_JAVA_OPTS property in the /etc/default/cloudera-scm-server file. The last line of the file is:

                  export CMF_JAVA_OPTS="-Xmx2G -XX:MaxPermSize=128m"
                

where -Xmx2G option specifies 2 GB of RAM. You can increase the setting, if necessary.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: Edit /etc/default/cloudera-scm-server as described above.

— (Applies to CDH4 only) Secure HDFS High Availability with Automatic Failover enabled does not use authorization against ZooKeeper.

This is a potential security issue because unauthorized users could manipulate the failover management state.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: None

— (Applies to CDH4 only) HDFS Security and Federation do not work together.

Only HBase works with the HDFS Security. The MRv1, MRv2 (YARN), Oozie, and Hue services do not work with HDFS Security.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: None

— The "corruptBlocks" metric actually shows corrupt replicas, not corrupt blocks.

The "corruptBlocks" metric, used by Cloudera Manager's corrupt blocks health check, actually returns the number of corrupt replicas, not the number of corrupt blocks. Therefore, if Cloudera Manager detects a corrupt replica it reports it as a corrupt block, even though the block itself is not actually corrupt. However, hadoop fsck correctly returns the number of corrupt blocks — meaning all replicas of those blocks are corrupt. Thus, Cloudera Manager may report a corrupt block (due to detecting a corrupt replica) while fsck will report that same block as healthy because one or more valid replicas still exist for the block.

Severity: Medium

Anticipated resolution: Fixed in Cloudera Manager 4.0.1

Workaround: You can use hadoop dsfadmin -report to verify the number of corrupt replicas, vs. hadoop fsck to verify the number of corrupt blocks.

— MapReduce jobs are sometimes marked as failed even though they are running and ultimately succeed.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.1

Workaround: None, but you can drill down into the tasks to show the correct state of running attempts for the MapReduce job.

— Cloudera Manager uses timezone of browser for displaying date/time, not timezone of the server.

In Cloudera Manager 4.0.x, the displayed date and time is now using the timezone of the Cloudera Manager server host, rather than the timezone of the user's browser.

Severity: Medium

Resolution: Fixed in Cloudera Manager 4.0.1.

Workaround: None