Skip to main content
Version: 1.0

Hive

  • Is data warehouse
  • Uses HQL

Services​

  1. Hive
  2. Metastore
    • Stores metadata of tables and partitions
    • Hive in remote mode:
      • Metastore use its own JVM process and other services communite with it using Thrift API
  3. HiveServer2

Tuning​

  • Memory configuration
  • GC limits
  • Partitions number limits

HA​

  • Configure load balancer for HiveServer2
  • Enable HA for metastore

Replication​

Security​

  1. Authentication: LDAP and Kerberos
    • If Kerberos is not enabled no need to provide password when connecting using Beeline CLI
  2. Authorization: Ranger
  3. Encryption SASL QOP or SSL/TLS

About​

  • Hive do not enforce schema on write, but it

hive-site.xml​

  • hive.metastore.warehouse.dir : default dir in hdfs where data is stored
  • Metastore
  • Driver: javax.jdo.option.ConnectionDriverName
  • Password: javax.jdo.option.ConnectionPassword
  • Url: javax.jdo.option.ConnectionURL

core-site.xml​

Data Model​

Table​

  • Show

    • show tables in <namespace>
    • show tables 'emp*'

  • Describe

    • describe extended <tablename>
    • describe formatted <tablename>

  • Describe column

    • describe extended <tablename.columnname>

  • Internal (managed)

    • Data stored in default location
    • The data is dropped when the table is dropped

  • External

    • Location is not the default one
    • Dropping table will not drop data, although table's metadata will be dropped

Partition​

  • strict: prohibit queries without where clause in partitioned tables

  • nonstrict: allow running queries without where clause

    set hive.mapred.mode=strict;

  • We can describe and show partitions

    • show partitions <tablename>
    • show partitions <tablename> PARTITION(col_name='value')

  • We can alter table by adding partition

  • no support for insert, update and delete

  • doesn't support transactions

      Create table if not exists mydb.employee(
          name string,
          salary float,
          adgress struct<street: string, city: string, state: string, zip: int> 
      )
      PARTITIONED BY (country string, state string);
    • Load data from file to partitioned table
    LOAD DATA LOCAL INPATH '${env:HOME}/path/to/dir'
    INTO TABLE employees
    PARTITION (col_part_name1 = 'value1', col_part_name2 = 'value2');

Bucket​

  • Each bucket is stored as file in partition directory

Types​

  • Java types (int, string, float, ...)
  • Complex types: Array, Map, Struct

Internal vs external table​

  • Hive manages interanl but non external
  • Drop internal delete metadata and path, but drop external remove only metadata
  • Internal support archive, unarchive, ACID/transactional and result caching ?? but none of there features are supported by an external