Deploying Greenplum

You are now ready to deploy Greenplum Database on the newly deployed cluster. Perform the steps below from the Greenplum master node.

Deploying a Greenplum Database Cluster

1. Initialize the Greenplum cluster.

   1. Log in to the Greenplum master node as gpadmin user.

   2. Create the Greenplum configuration script create_gpinitsystem_config.sh and paste the following contents:

      #!/bin/bash
      # setup the gpinitsystem config
      primaryArray() {
        numOfSegments=$1
        array=""
        newline=$'\n'
        for i in $(seq 1 ${numOfSegments}); do
          array+="sdw$(($i*2-1))~sdw$(($i*2-1))~6000~/gpdata/primary/gpseg$(($i-1))~$(($i*2))~$(($i-1))${newline}"
        done
        echo "${array}"
      }
      mirrorArray() {
        numOfSegments=$1
        array=""
        newline=$'\n'
        for i in $(seq 1 ${numOfSegments}); do
          array+="sdw$(($i*2))~sdw$(($i*2))~7000~/gpdata/mirror/gpseg$(($i-1))~$(($i*2+1))~$(($i-1))${newline}"
        done
        echo "${array}"
      }
      create_gpinitsystem_config() {
      echo "Generate gpinitsystem"
      cat <<EOF> ./gpinitsystem_config
      ARRAY_NAME="Greenplum Data Platform"
      TRUSTED_SHELL=ssh
      CHECK_POINT_SEGMENTS=8
      ENCODING=UNICODE
      SEG_PREFIX=gpseg
      HEAP_CHECKSUM=on
      HBA_HOSTNAMES=0
      QD_PRIMARY_ARRAY=mdw~mdw~5432~/gpdata/master/gpseg-1~1~-1
      numTotalSegments=$1
      declare -a PRIMARY_ARRAY=(
      $( primaryArray $((${numTotalSegments}/2)) )
      )
      declare -a MIRROR_ARRAY=(
      $( mirrorArray $((${numTotalSegments}/2)) )
      )
      EOF
      }
      numTotalSegments=$1
      if [ -z "$numTotalSegments" ]; then
        echo "Usage: bash create_gpinitsystem_config.sh <num_total_segments>"
      else
        create_gpinitsystem_config ${numTotalSegments}
      fi
      

   3. Run the script to generate the configuration file for gpinitsystem. Replace 64 with the number of segments in your environment:

      $ bash create_gpinitsystem_config.sh 64
      

      You should now see a file called gpinitsystem_config.

   4. Run the following command to initialize the Greenplum Database:

      $ gpinitsystem -a -I gpinitsystem_config -s smdw
      

2. Configure the Greenplum master and standby master environment variables, and load the master variables:

   $ echo export MASTER_DATA_DIRECTORY=/gpdata/master/gpseg-1 >> ~/.bashrc
   $ ssh smdw 'echo export MASTER_DATA_DIRECTORY=/gpdata/master/gpseg-1 >> ~/.bashrc'
   $ source ~/.bashrc
   

3. Configure the Greenplum cluster with the commands below. Note that some of the parameter values will vary, depending on your virtual machine RAM size.

   ### Interconnect Settings
   $ gpconfig -c gp_interconnect_queue_depth -v 16 
   $ gpconfig -c gp_interconnect_snd_queue_depth -v 16
   
   # Since you have one segment per VM and less competing workloads per VM,
   # you can set the memory limit for resource group higher than the default 
   $ gpconfig -c gp_resource_group_memory_limit -v 0.85 
   
   # This value should be 5% of the total RAM on the VM
   $ gpconfig -c statement_mem -v 1536MB 
   
   # This value should be set to 25% of the total RAM on the VM
   $ gpconfig -c max_statement_mem -v 7680MB 
   
   # This value should be set to 85% of the total RAM on the VM
   $ gpconfig -c gp_vmem_protect_limit -v 26112
   
   # Since you have less I/O bandwidth, you can turn this parameter on
   $ gpconfig -c gp_workfile_compression -v on 
   

4. Restart the Greenplum cluster for the newly configured settings to take effect:

   $ gpstop -r
   

Validating the Greenplum Installation

Run the commands below from the master node as gpadmin user in order to validate basic functionality of the Greenplum cluster.

1. Refresh the environment variables:

   $ source ~/.bashrc
   

2. Show the state of the Greenplum cluster:

   $ gpstate
   

3. Connect to the postgres database and check segment configuration information:

   $ psql postgres
   postgres=# SELECT * FROM gp_segment_configuration ORDER BY hostname;
   

4. Once connected to the postgres database, verify that you can create a table, insert data to it and read it:

   postgres=# CREATE TABLE t AS SELECT generate_series(1, 1000000);
   postgres=# SELECT MIN(cnt), MAX(cnt), COUNT(cnt) FROM (SELECT COUNT(*) cnt FROM t GROUP BY gp_segment_id) tt;
   postgres=# DROP TABLE t;