Known Issues Fixed in Cloudera Manager 4.5.1

— Under HDFS HA configuration, recommissioning a DataNode requires both active and standby NameNodes to be running.

In an HA-enabled HDFS cluster managed by Cloudera Manager, recommissioning a DataNode when the active NameNode is started but the standby NameNode is stopped will fail because dfsadmin -refreshNodes will fail on the standby NameNode (which isn't started).

Severity: Low

Anticipated Resolution: Fixed in Cloudera Manager 4.5.1

Workaround: None.

— HDFS replication and Hive replication fails on a cluster with CDH version less than 4.2 if HDFS has data transfer encryption enabled.

HDFS replication and Hive replication (with the "Replicate HDFS Files" option enabled) fails on a cluster with CDH version less than 4.2 if HDFS has data transfer encryption enabled, that is, if dfs.encrypt.data.transfer is set to true.

Severity: Medium

Resolution: Fixed in Cloudera Manager 4.5.1

Workaround: Disable checksum checks for the replication. This can be done by enabling the "Skip Checksum Checks" option for the replication schedule.

— Upgrade of the Event server fails if /var runs out of disk space.

Severity: Medium

Resolution: Fixed in Cloudera Manager 4.5.1

Workaround: None.

— (Applies to Oracle only)   Installation fails after the "Review Configuration Changes" step in the Wizard.

On a cluster running the Oracle OS, installing on a large number of hosts in a single batch, the installation may fail and generate a stack trace when you click "Continue" after the "Review Configuration Changes" step. The error you will see in the log is ORA-01795: maximum number of expressions in a list is 1000. This will occur when the total number of roles being added is greater than 1000. The number of roles is a function of the number of hosts and the number of roles created across those hosts for the services you are adding. (For example, the HDFS service by default creates a DataNode role on every host).

Severity: High

Resolution: Fixed in Cloudera Manager 4.5.1.

Workaround: Split your installation into batches — either install on fewer hosts initially, and then add additional hosts later, or install fewer services initially and add the other services later.

— If no Hive Metastore Server roles are defined or they have been deleted, various Hive service pages give a popup with an error message.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.5.1.

Workaround: Dismiss popup whenever it appears. Add a Hive Metastore Server instance.

— If you modify multiple group parameters in a Cgroup configuration, they do not update dynamically

Cloudera Manager 4.5 introduces support for the Control Groups (cgroups) Linux kernel feature. With cgroups, administrators can impose per-resource restrictions and limits on CDH processes. If you modify multiple group params at once, the dynamic refreshing of cgroups fails.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.5.1.

Workaround: Make Cgroup-based resource management parameter changes one-by-one. After each change, save the results before making further changes.

— JobTracker and TaskTracker RPC and UGI metrics not supported

RPC and UGI metrics for Job and Task Tracker that were available in CDH3 are not available due to changes in CDH4.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.5.1.

Workaround: None.

— HBase RegionServers cannot be decommissioned if using Federated HDFS.

HBase RegionServers cannot be decommissioned if their HDFS dependency is federated. This is because the commands to move regions off RegionServers fail with viewfs. This will be fixed in a future CDH release.

Severity: Low

Resolution: Fixed in CDH4.2

Workaround: None.