Upgrade requirements

Before upgrading, make sure your system meets the following requirements.

Requirement Description
Hadoop BDD 1.3.2 supports the following:
  • Cloudera Distribution for Hadoop (CDH) 5.5.x (min. 5.5.2), 5.6, 5.7.x (min. 5.7.1), 5.8
  • Hortonworks Data Platform (HDP) 2.3.4.17-5, 2.4.x (min. 2.4.2)
  • MapR Converged Data Platform (MapR) 5.1

If you don't have one of the above installed, upgrade your Hadoop cluster before upgrading BDD. For instructions, refer to the documentation for your Hadoop distribution.

Before upgrading your Hadoop cluster, be aware of the following:
  • You can't switch to a different Hadoop distribution without reinstalling BDD. For example, if you currently have CDH, you can't switch to HDP.
  • You should stop your BDD cluster before you upgrade Hadoop. Once Hadoop has been upgraded, follow the procedure described in "Switching Hadoop versions" in the Administrator's Guide to enable BDD to work with the new version.
  • After you upgrade Hadoop, verify that the YARN configuration changes you made before installing BDD weren't reset. For more information, see "YARN setting changes" in the Installation Guide.

If you're upgrading from BDD 1.0, upgrade Hadoop just before you run the BDD upgrade script. BDD 1.0 doesn't support any of the currently-supported versions of CDH, so your BDD cluster must remain stopped once you upgrade Hadoop. Additionally, you should verify that the server roles in your Hadoop cluster don't change during the upgrade.

Secure Hadoop BDD 1.3.2 can run on Hadoop clusters secured with TLS/SSL and HDFS data at rest encryption. If you don't currently have these enabled, you can configure them in your Hadoop cluster before upgrading BDD and then enable them for BDD at upgrade time.

See the section "Security options" in the Installation Guide for instructions. Also, be sure to set the HADOOP_CERTIFICATION_PATH property in bdd.conf before upgrading.

Operating system BDD supports the following operating systems:
  • Oracle Enterprise Linux 6.4+, 7.1
  • Red Hat Enterprise Linux 6.4+, 7.1

If you don't have one of the above installed, upgrade your OS before upgrading BDD.

Additionally, you must clear the $http_proxy environment variable: 
export http_proxy=
OS user The entire upgrade process must be performed by the bdd user (the OS user that installed BDD and runs all BDD processes). This user must have the same UID on all nodes in the cluster, including Hadoop nodes.
Dgraph databases (index) The Dgraph database (index) directory must contain enough free space to double your databases, as this may occur temporarily during the upgrade process.
Note: Although BDD 1.3.2 enables you to store your Dgraph databases (index) on HDFS, you can't move them there until after the upgrade. For instructions on doing this, see the Administrator's Guide.
Transform Service If you're upgrading from 1.0 or 1.1.x, the upgrade script will install the Transform Service on the nodes you specify in bdd.conf. For best performance, these should be WebLogic Managed Servers.

The Transform Service requires at least 11GB of RAM and 39GB of virtual memory, and may require more depending on the size of its workloads. Be sure that the nodes you install it on meet these requirements.

Perl modules Verify that the following Perl modules are installed on the Admin Server:
  • Mail::Address
  • XML::Parser
  • JSON-2.90

If any are missing, install them according to the instructions in "Installing the required Perl modules" in the Installation Guide.

Linux utilities curl 7.19.7+ and Network Security Services (NSS) 3.16.1+ must be installed on all Studio nodes. Additionally, curl must support the options --tlsv1.2 and --negotiate.