Known Issues Fixed in Cloudera Manager 4.0.4

— RegionServer heap size override doesn't have any effect in a Kerberos environment.

When running HDFS, HBase, and ZooKeeper in a Kerberos environment, setting a different value for the RegionServer Heap size was not passed along to the jvm where the role instances is running.

Severity: Medium

Resolution: Fixed in Cloudera Manager 4.0.4

— ulimits are not set for the Cloudera Manager server.

Severity: Medium

Resolution: Fixed in Cloudera Manager 4.0.4

Workaround: Create a file called /etc/security/limits.d/cloudera-scm.conf with the following contents:

cloudera-scm - nofile 32768 cloudera-scm - nproc 65536

— Cloudera Manager is reporting Bad Health due to worker node connectivity issues.

This may happen when the master node's view of the worker node's name differs from Cloudera Manager's view of the name. Cloudera Manager has been made more robust in name matching in such situations.

Severity: Medium

Resolution: Fixed in Cloudera Manager 4.0.4

— Logging in as LDAP user may cause a stack trace.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.4

Workaround: Uncheck the "Remember me" option in the browser

— Running a custom version of SSH does not work with Cloudera Manager upgrade

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.4

— The Services Monitoring process may misreport the health of HDFS in a certain situation.

In a HA configuration, the Services Monitoring process may misreport the health of HDFS in the following situation: if the process can communicate with the active NameNode, but not with the standby NameNode, it may misreport that the standby NameNode is not running. Instead, it should indicate that it cannot communicate with the standby NameNode. This known issue is only about the wording of the message; the health check correctly returns a Concerning health.

Severity:* Low

Resolution: Fixed in Cloudera Manager 4.0.4

Workaround: None

— After upgrading from Cloudera Manager 3.7, HDFS health check metrics are not available.

After upgrading from a 3.7.x version of Cloudera Manager, health check metrics for HDFS and MapReduce services report "Not enough data...." errors. This is due to a missing property setting.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.4

Workaround: Restarting the affected roles will regenerate the properties files, which will now contain the missing properties.

— Monitoring does not work properly if you select LDAP Only for the Authentication Backend Order property.

Monitoring does not work properly if you select LDAP Only for the Authentication Backend Order property in the Administration > Properties tab of the Cloudera Manager Admin Console.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.4

Workaround: If you want to use LDAP authentication, select LDAP then Database for the Authentication Backend Order property, and delete all extraneous users on the Administration > Users tab.

— Unable to authenticate against an OpenLDAP compatible server.

Search filters were added in Cloudera Manager 4.0.3, but the required groups filter option wasn't documented. This has been added in the Cloudera Manager User Guide.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.3. Documentation enhanced in Cloudera Manager 4.0.4.

— /var filling up due to Reports Manager not releasing space after deleting files.

The Reports Manager is not releasing file descriptors (and the related disk space) after building its index.

Severity: Low

Resolution: Fixed in Cloudera Manager 4.0.4

Workaround: Restart the Reports Manager to free the file descriptors.

— Event store cleanup is deleting the wrong events.

When the event store gets full, Cloudera Manager deletes some portion of the events, starting with the oldest events. However, the cleanup of the event store was erroneously marking events that are not the oldest for deletion, instead of deleting the oldest events.

Severity: Medium

Resolution: Fixed in 4.0.4

Workaround: None