I am new to Teradata (spooling up POC) and have thus far not been able to locate the infomation which I am seeking so beforehand I do apologize for such basic questions, but I would greatly appreciate some assistance.
1. How do we setup connectivity to Teradata 4-node 2690 appliance? That is, do we need to setup DNS VIP for round-robin across all nodes or does Teradata handle this somehow (Query Director perhaps)?
2. Once a user connects to a node, their queries are regardless run across all nodes, corect? Is this also the case with data ingest which would indicate the ingest rate is limited only by the number of nodes (where each node has active/passive 10 GigE NIC for ~1.2 GB/s or ~4.2 TB/hr throughput rate per node)?
3. If a node fails during backup/recovery, is the entire backup/recovery process failed or just the session running on the failed node which gets restarted on another available node (HSN)?
I appreciate your time and kind assistance.
1) Connectivity can be set up to some or all nodes in the system using DNS. Teradata handles the round-robing of sessions across all nodes that are set up in the DNS provided they are set up correctly. All of this is seamless to the end user who simply provides a TDP ID, user ID, and password to connect to Teradata. There is no specification of a specific node or nodes done by the user.
2) Although a user connects through a particular node, their queries access the nodes containing the data. Typical decision support queries will access multiple or all nodes, while tactical type queries may only access one. It depends on the tables inolved, their primary index definitions, and the query that it submitted. Teradata automatically takes care of accessing one/multiple/all nodes as required to solve the query.
3) If a node fails, a Teradata restart is performed. If the node does not come up as part of the restart, it is left out of the configuration and the sessions/AMPs that were on that node are migrated to the other nodes in the clique. Normally when this happens, the client software automatically submit the queries that were in process and the load/export jobs can automatically continue from where they left off.