This topic describes new features in Cloudera Manager.
Core Configuration Service🔗
The Core Configuration service allows you
to create more types of clusters without having to include the HDFS
service. Previously, the HDFS service was required in many cases even when
data was not being stored in HDFS because some services like Sentry and
Spark required cluster-wide configuration files that Cloudera Manager
deploys within the HDFS service. The Core Configuration service provides
this configuration in a standalone fashion and thus eliminates the need
for an HDFS service for certain types of clusters where no HDFS storage is
required (e.g. Kudu, Kafka, or ‘Compute’ clusters using exclusively object
storage like S3 or ADLS). The Core Configuration service is also useful
when creating a Compute cluster that accesses data on an HDFS service
located in the Base cluster.
"Impala for Compute" and "Spark for Compute" no longer require HDFS.
You can define the Core Configuration Service instead.
Metric Filtering🔗
Metrics Filters allow you to
limit the amount of metric data sent to the Cloudera Manager Service
Monitor. In large clusters, some services, such as Kudu, send a high
volume of non-essential metrics data to the Service Monitor, which can
overload it, causing gaps in the data reported from these metrics in
charts, dashboards, and metric queries, and potentially limiting the
ability for Cloudera Manager to effectively monitor cluster health. To
mitigate this problem, you can configure Metric Filters that
limit the amount of data sent to the Service Monitor. You can configure
Metric Filters for any service deployed in a cluster.
See Filtering Metrics
YARN Scheduler🔗
The default scheduler in the YARN Resource Manager is set to Capacity
Scheduler. Preemption, Node Labels, Queue Mutation API, and Asynchronous
Scheduling are enabled by default.
Solr, HBase, and Kudu on Compute Clusters🔗
Creation of Solr, HBase and Kudu services on Compute Clusters is now
enabled.
LDAP authentication for Kafka clients🔗
You can now configure LDAP to allow Kafka clients to authenticate
using LDAP.
Auto TLS🔗
You can now use Cloudera Manager to Automatically configure TLS for
your clusters.
HTTP Strict-Transport-Security🔗
When TLS is enabled for the Cloudera Manager Admin Console, web
requests now include the HTTP Strict-Transport-Security header. For more
details about this header, see Strict-Transport-Security
(Mozilla).
Diagnostic Bundle Changes🔗
The value of the Cloudera Manager parameter CDP_ENVIRONMENT now appears
in support bundles from Data Center deployments.
Ranger Service and Kafka🔗
The Ranger service name for Kafka clusters is now configurable. The
default (and initialized) value is cm_kafka
.
Cloudera Manager Licensing🔗
When a license for Cloudera Manager expires, or the trial period
expires, access to the Cloudera Manager Admin Console will be limited to
only the license page until you install a new valid license. The
Cloudera Manager Admin Console features will no longer be disabled, but
you will be unable to view or modify those features from the Cloudera
Manager Admin Console.
New Health Tests 🔗
- LDAP connections. The LDAP health check requires you to set a bind
user to enable monitoring.
- Key Distribution Center (KDC) connections. The KDC health check
requires Cloudera Manager Server to use Kerberos to enable
monitoring.
New configuration parameters for Azure🔗
Two new core-site configurations have been added to support delegation
token collection on Azure cloud storage:
- fs.azure.identity.transformer.service.principal.substitution.list
- fs.azure.identity.transformer.service.principal.id
New Kafka Metric🔗
A new metric has been added to the Kafka service for JVM Garbage
Collection Rate: kafka_jvm_gc_runs
.
New notification suppression parameters🔗
Notification suppression parameters for role-level validators are now
available.
Redaction in Cloudera Manager API🔗
Previously redaction was opt-in through a JVM parameter, causing major
security concerns. Customers relying on the API for backups now have a
viable alternative that does not rely on exposing passwords via the API.