What's New in Cloudera Manager 4.5.0

  • Rolling upgrades. You can now manage and upgrade the CDH binaries on the cluster using Cloudera Manager with minimal downtime. This feature supports a new packaging format called "parcel" to facilitate the upgrade mechanism. You can now download, distribute, and activate a new CDH version parcel all from within Cloudera Manager. See Managing Parcels. If you have HDFS High Availability enabled, you can perform a rolling upgrade, which does a parcel update and then offers to do a cluster-level rolling restart. See Performing a Rolling Upgrade on your Cluster. Parcels are available for CDH 4.1.3 and onwards. Cloudera Manager continues to work with RPM (RHEL and SLES) and Debian/Ubuntu packages.

  • Rolling restarts. Restart services with zero/minimal downtime as related to upgrades (new versions of parcels) or configuration changes. You can do a rolling restart on individual services (HDFS, MapReduce, HBase, ZooKeeper, and Flume services). Rolling restart also lets you choose which roles to restart. If you have HDFS High Availability enabled, you can do a rolling restart on the entire cluster.

  • Improved configuration management for role instances. You can now group an arbitrary set of roles instances (of the same type, for example, DataNodes) together and manage their configuration as a unit. Called "Role Groups", this feature simplifies configuring groups of roles that need to different configuration settings, for example when running on nodes with heterogeneous hardware.

  • Improved role to host assignment with Host Templates. A host template defines a set of role configurations that can be applied to new hosts. This simplifies the process of adding hosts and instantiating the roles that should run on those hosts.

  • Resource management. Enables partitioning of cluster resources (CPU, memory) across different services in use. Enables customers to run multiple workloads/services simultaneously (for example, MapReduce and Impala on the same cluster) with no loss in QoS.

  • Advanced charting. Provides intuitive and customizable charts to display relevant time-series metrics across the board (services, role instances, hosts). Includes a new "Charts" feature to create (and save) dashboards of metrics. Saved views of defined charts enable quick diagnostics/trouble shooting. Helps better correlate metrics from services to roles to hosts.

  • SNMP support. Makes alerts available as SNMP traps which enables easier integration with IT management tools such as IBM Tivoli, HP Openview, and others.

  • Automated cluster statistics. Scheduled, periodic diagnostics gathering (enabled as default) and reporting now enable Cloudera Support to proactively address customer issues. The diagnostic function collects configuration files, node count, software versions, log files and other information regarding customer’s environment and reports that information to Cloudera to diagnose any potential issues. Customers may elect to change configuration to disable regular automatic reporting or to report only on filing of a support ticket. Also see the Cloudera Privacy Policy.

  • HBase monitoring improvements. Provides new health checks, metrics, events (compactions, splits, and so on), periodic data integrity checks (hbck), and region liveness checks (HBase canary).

  • Backup and Disaster Recovery. An independently-licensed feature that enables you to replicate HDFS data files and Hive metastore and data files from one Cloudera Manager-managed cluster to another cluster managed by a different Cloudera Manager server. You can configure this on a one-time or on-demand basis or on a recurring schedule.

  • Support for Hive. Cloudera Manager can automatically set up the Hive Metastore for you. It also supports running the Hive Metastore Server and HiveServer2. If you have configured Hue or Impala, then upgrading to Cloudera Manager 4.5 will automatically create new Hive service(s), migrate the existing Hive related configuration over, and link Hue and Impala to the new Hive service(s).

  • Support for HBase REST and Thrift servers. These can be added as role instances for HBase services.

  • AWS installation wizard. When running on AWS (Amazon Web Services), the Cloudera Manager installation wizard helps you bring up new EC2 nodes to form the cluster, and optimize the settings for the AWS environment. This provides an expedient way to create a fully functional CDH cluster managed by Cloudera Manager, although it is not intended for long-running production use.

  • Hue dependency on MapReduce. In CDH3 clusters, Hue had a dependency on the MapReduce service. Starting with Cloudera 4.5, Hue has a dependency on the Hive service, which indirectly satisfies the MapReduce dependency.

  • Automatic creation of Oozie ShareLib. See CDH4 Installation Guide. Oozie ShareLib is automatically installed when adding an Oozie service.

  • HDFS Short-circuit reads improvement. In Cloudera Manager 4.1, HDFS short-circuit reads is enabled only for Impala daemons. In Cloudera Manager 4.5, HDFS will perform short-circuit reads if clients are so configured. By default, Impalad and HBase RegionServers enable short-circuit reads, while MapReduce TaskTrackers and all gateways (HDFS, MapReduce, HBase) disable short-circuit reads.

  • Several usability enhancements. Includes improved "Service" pages, "Heatmaps", and event search.