How to initialize Greenplum on multiple nodes

Actual situation

If tou have followed previous article, you have a Greenplum installed on multiple nodes.

Standar procedure when dealing with a Greenplum database, as well as Postgres, is composed by:

• Installation
• Initialization
• Database Start

In this article, we will see the second and third steps: initializating and starting the database.

Database initialization

The script that do the job here is gpinitsystem. gpinitsystem needs a special configuration file, who contains a list of segment host addresses only. Let's name it hostfile_gpinitsystem.

For example:

segment-hostname-1
segment-hostname-2
...

One more file is needed, its name is gpinitsystemconfig. It contains a lot of parameters to configure your system. An example configration file, to be used as a template, is in $GPHOME/docs/clihelp/gpconfigs/gpinitsystem_config. You can copy that example file and modify it to suits your needs. A detailed list of all parameter meanings is on Admin Guide at page 67.

The very important part of the file is:

ARRAY_NAME="EMC Greenplum DW"
PORT_BASE=40000
declare -a DATA_DIRECTORY=(/data1/primary /data1/primary
/data1/primary /data2/primary /data2/primary /data2/primary)
MASTER_HOSTNAME=master-hostname
MASTER_DIRECTORY=/data/master
MASTER_PORT=5432

Now you can run the gpinitsystem utility with those two files as parameters, this way:

$ gpinitsystem -c gpinitsystem_config -h hostfile_gpinitsystem

After a succesfully initialization you will see a kind message:

Greenplum Database instance successfully create+.

In case of failure, it is possible to get a partially installed system.

Because of that, after a failure in gpinitsystem Greenplum automatically creates a "rollback" script that can be executed to cleanup things around.

An example cleanup script file would be named:

backout_gpinitsystem_gpadmin_20111216_121053

Is it possible to execute that using GNU bash:

$ sh backout_gpinitsystem_gpadmin_20111216_121053

All Greenplum files partially installed will be removed.

That's all for now. I hope that this article, together with the previous one, will show you how easy Greenplum installation is.