What's New in Cloudera Manager 7.4.4
New features and changed behavior for Cloudera Manager 7.4.4.
- Upgrades from CDH 6 to CDP are now supported.
- You can upgrade from CDH 6.1 or higher to CDP Private Cloud Base. See In-place upgrade from CDH 6 to CDP Private Cloud Base.
- New Support for Dell OneFS (Isilon)
- The following upgrades are now supported for Dell EMC PowerScale OneFS:
- CDH 5 to CDP Private Cloud Base 7.1.7
- HDP 2 to CDP Private Cloud Base 7.1.7
- New installations on Dell OneFS are now supported for CDP Private Cloud Base 7.1.7.
- New Upgrade Guide Companion app
We are introducing a companion to the already existing Upgrade Guide we currently publish with all releases. The Upgrade Companion is a stand-alone web page that is hosted as part of docs.cloudera.com but is separate from the rest of the product documentation. It will provide a central hub for all activities related to upgrading our products regardless of the form factor. The Upgrade Companion will provide a set of content tabs that cover the major points of a customer's upgrade journey:
- Getting Started
- Pre-upgrade Tasks
- Upgrade the Cluster
- Post-upgrade Tasks
Each tab includes content and links that help the user navigate that particular step in the journey. In addition, the Upgrade Companion will also include a full set of filters that are used to filter content based on specific configurations selected.
- New supported operating systems
- The following operating systems are now supported for use with CDP
Private Cloud Base 7.1.7:
- RHEL 8.2
- Ubuntu 20
- increase replication policies page refresh rate greater than 15sec
- The auto-refresh interval on the Replication and Snapshots pages has been increased from 15 seconds to 2 minutes. A refresh button has been added to allow more frequent refreshes if desired;
- New supported databases
- The following databases are now supported for use with CDP Private
Cloud Base 7.1.7:
- MariaDB 10.3 and 10.4
- Oracle 19.9
- Rollback of upgrades to CDP
- Rollback is now supported for the following upgrades:
- HDP 3 to CDP 7.1.7 (Cloudera Manager 7.4.4) or higher. See Rollback HDP Services from CDP 7.1.7
- CDH 6 to CDP 7.1.7 (Cloudera Manager 7.4.4) or higher. See Rolling back a CDH 6 to CDP Private Cloud Base upgrade for the procedures.
- Add test LDAP Configuration feature in Cloudera Manager
- The configured LDAP settings can now be tested
through the Cloudera Manager Admin Console or API. After saving, LDAP
settings can be tested without restarting the Cloudera Manager server.
This is available in the Cloudera Manager Admin Console under
Administration -> Users & Roles -> LDAP/PAM Groups and in the API
- Cloudera Manager agent to support static UID+GID generation natively
- Cloudera Manager now allocates fixed numeric
IDs for service user accounts and service group accounts (for example,
the HDFS service runs in the "hdfs" user account).
Previously, Cloudera Manager allowed the OS on each host to choose the numeric ID underlying the service accounts (for example, service user account "hdfs" would be allocated numeric ID 986). This resulted in different hosts having random assignments of numeric IDs, which would make it problematic to move partitions between hosts.
With this release, Cloudera Manager statically allocates numeric IDs when a new host is added to Cloudera Manager. For services in the CDP parcels, the numeric ID will come from a static list. For services in other parcels, the numeric ID is chosen by hashing the service name.
If a service user account or group account already exists, Cloudera Manager will not change the numeric ID. Therefore existing hosts will not be changed. If customers were previously depending on the semi-random allocation of numeric IDs to match between hosts, this change will cause new hosts to have a different (though static) allocation from previously added hosts.
If customers pre-allocate service user and group accounts for the Cloudera Manager managed services, Cloudera Manager will not modify those accounts.
The feature is turned on by default. To disable it, customers should set the property
is_static_uid_gid_enabledto false in
/etc/cloudera-scm-agent/config.inion all cluster hosts.
- New configuration property for Streams Messaging Manager CSD
- The property
prometheus.query.threadsdetermines how many parallel requests Streams Messaging Manager should send to Prometheus when fetching metrics (in case Prometheus is used as a backend metrics store). This value should match the value of the
query.max-concurrencyproperty in Prometheus. The default value for both is 20.0.
- New Date Format for killed Java process message
- When a Java process is killed due to
exceeding memory resources (Out of Memory - OOM), Cloudera Manager
creates a file called
killed_by_killparent_on_oomin the process directory. The timestamp stored in that file has been updated to use ISO 8601 format.
- Make the YARN/HDFS SSL/TLS Cipher Suite's value dependent on FIPS compliant mode
- For FIPS compliant clusters, the SSL/TLS Cipher Suite should be set to Intermediate 2018 in order to be able to access the web UIs of the services. Hive, HBase, HDFS and YARN already use Intermediate 2018 as the default for FIPS mode.
stacks_collection_directorywill also set ownership of the directory to the process user
- Cloudera Manager now displays a warning message when users try to set the "Stacks Collection Directory/ Heap Dump Directory" to non-recommended paths. Previously, the operation of setting "Stacks Collection Directory/ Heap Dump Directory" could change the ownership of any system directory. Also, sharing the same directory among multiple service roles caused an ownership race. Thus, any process that tries to access the folder might be denied and fail to start.
- With this release, Cloudera Manager guides the
user to set those paths under
/var/log. (For example:
/tmp/hdfs_stacks_collection_dirs). If this rule is contradictory to a previous deployment, you can suppress or ignore the warning message.
- Cloudera Manager now configures Atlas Hooks for Sqoop
- Cloudera Manager is now capable of enabling
and configuring the Sqoop Atlas Hook. See: http://atlas.apache.org/2.0.0/Hook-Sqoop.html. The configuration happens automatically, no
extra, manual configuration is required from the users. Cloudera
Manager will configure the Hook in sqoop-site.xml and it will
automatically generate the
atlas-application.propertiesfile for Sqoop. If you are installing a fresh cluster with Atlas being present, then the Hook will be enabled automatically and the
atlas-application.propertieswill be generated for Sqoop. If you are upgrading from an older cluster then you need to enable the Hook manually. To enable Hook manually:
- Go to the configuration page for the Sqoop service.
- Search for "Atlas".
- Enable the checkbox and then re-deploy the client configurations using Cloudera Manager.
- : Add new auth-to-local rule for Atlas
- Support has been added for adding an auth-to-local rule for Atlas principal to Atlas service user mapping.
- Improved start up performance of YARN and MapReduce jobs.
- Workload setup for YARN, MapReduce is now about 20% faster.
- Cloudera Manager changed topology.map file generation from XML to INI format to improve the performance of the script topology.py. topology.py has been modified to process topology.map in INI format.