The Hadoop cluster has multiple network connections. Each node in the cluster has a minimum of three physical connections:
- Server Management
- 1GbE primary and secondary connections
- 40Gb/sec redundant connections used for the Hadoop cluster and connectivity to edge nodes
- Public Interfaces
- 1GbE or 10GbE external interfaces used to access customer network resources and for inbound customer client requests to the cluster
Hadoop communicates on the InfiniBand internal network.
The corporate DNS database must include the fully-qualified Hadoop node names.
Teradata installs and configures networking in the system, but the public interfaces can be customized.
Cloudera provides proactive support to prevent misconfiguration and improve troubleshooting. To automatically send diagnostic diagnostic bundles to Cloudera, Master Node 1 must have external Internet access. If Master Node 1 cannot be given external Internet access, data bundles can be sent manually, but there will be no proactive support. See https://www.cloudera.com/documentation/enterprise/latest/topics/cm_ag_data_collection.html#cmug_topic_15_4
for more information.
Cloudera Manager uses Google Analytics to send anonymous usage information. By default, anonymous usage data collection is enabled for diagnostic data bundles. If you have security concerns, open an incident with Teradata Support.
For setting up external connectivity to a customer LAN or external edge node, see Setting Up External Access.