CAST Storage Service - PostgreSQL 10 or above deployment on Linux

Installation via YUM 

PostgreSQL can be installed using RPMs (binary) or SRPMs (source) managed by YUM. This is available for the following Linux distributions (CAST recommends using only 64-bit platforms):

• Fedora
• CentOS
• Red Hat Enterprise Linux

Configure your YUM repository

Locate and edit your distribution .repo file, located:

• On Fedora: /etc/yum.repos.d/fedora.repo and /etc/yum.repos.d/fedora-updates.repo, [fedora] sections
• On CentOS: /etc/yum.repos.d/CentOS-Base.repo, [base] and [updates] sections
• On Red Hat: /etc/yum/pluginconf.d/rhnplugin.conf [main] section

To the section(s) identified above, you need to append a line (otherwise dependencies might resolve to the postgresql supplied by the base repository):

exclude=postgresql*

Update the RPM package

A PGDG file is available for each distribution. Browse https://yum.postgresql.org/repopackages.php and find your correct RPM. There is only one single repo RPM for all PostgreSQL versions for each platform. Ensure you modify the link to the rpm file in the command below: this command will download the RPM package for the latest release of Red Hat:

rpm -ivh https://download.postgresql.org/pub/repos/yum/reporpms/EL-8-x86_64/pgdg-redhat-repo-latest.noarch.rpm
yum update

Install PostgreSQL

First run the following command to view the available releases:

yum list postgresql*

Then run the install, for example to install PostgreSQL 11:

yum install postgresql11-server

Configure data storage location

The default location for PostgreSQL data storage is set to (where <name> is the release number):

/var/lib/pgsql/<name>/data

To customize this location, you need to edit the following file (where <name> is the release number):

/usr/lib/systemd/system/postgresql-<name>.service

Initialize the PostgreSQL database server to configure data storage location

Once installed, please run the following command to initialize the server (where <name> is the release number):

export PGSETUP_INITDB_OPTIONS="-E 'UTF-8' --no-locale"
/usr/pgsql-<name>/bin/postgresql-<name>-setup initdb

Configuring the server/database

Edit the pg_hba.conf file

Edit the pg_hba.conf file at /usr/local/pgsql/data using the vi command:

vi pg_hba.conf

Replace the sample 192.168.x.x IP address and /24 subnet below with the appropriate allowed client IP addresses/subnets:

# TYPE DATABASE USER CIDR-ADDRESS METHOD
local all all trust
# IPv4 local connections:
host all all 127.0.0.1/32 trust
host all all 192.168.21.57/24 md5
 # IPv6 local connections:
host all all ::1/128 md5

Note that adding the following line will ensure ANY client can connect:

host all all 0.0.0.0/0 md5

Check server configuration

To check if the server itself is configured correctly for CAST AIP, we need to verify and modify certain parameters in the postgresql.conf (the file is in the postgresql directory \db_data).

pgtune

The postgresql.conf settings listed below are directly related to available RAM for the PostgreSQL instance - and as such their configuration is specific to the host machine. Please use http://pgtune.leopard.in.ua/ on the host machine to identify the correct settings and then modify postgresql.conf with the settings provided by pgtune. You should update the following settings with the values provided by pgtune:

• effective_cache_size
• min_wal_size
• max_wal_size
• wal_buffers

Some of the values suggested by pgtune are somewhat low. Therefore, please calculate the following postgresql.conf settings as follows:

• shared_buffers - value should be 25% of available RAM with a max of 8 GB

• maintenance_work_mem= 512 MB

Other settings

The following settings should also be modified to the values listed below:

Start the PostgreSQL database server

Where <name> is the release number:

systemctl start postgresql-<name>.service
systemctl enable postgresql-<name>.service
systemctl status postgresql-<name>.service

Create users

By default PostgreSQL will create a system account user named postgres with the role postgres. The equivalent CAST Storage Service provided by CAST includes two default users as follows:

If you would like to create these users, use the following commands:

psql
*create user operator with SUPERUSER password 'CastAIP';
*create user guest with password 'WelcomeToAIP';
*grant postgres to operator;

CAST AIP ≥ 8.3.12 - custom users

If you are using CAST AIP ≥ 8.3.12, you are also free to create your own users and then use them with CAST AIP and related software - you need to have a minimum of one user with the SUPERUSER permission and then grant the "postgres" role to this user. For example:

[postgres@cssx data]$ psql
*create user my_user with SUPERUSER password 'my_password';
*grant postgres to my_user;

When using CAST AIP ≥ 8.3.30, it is possible to create custom users that do NOT require the SUPERUSER permission if you prefer not to grant this. For example, this script creates a role "my_role" with the password "my_password" that can login. The role has not been granted the SUPERUSER permission and instead only the minimum permissions required to operate CAST AIP are granted:

[postgres@cssx data]$ psql
*create role my_role LOGIN PASSWORD 'my_password' NOSUPERUSER INHERIT NOCREATEDB NOCREATEROLE NOREPLICATION;
*grant create, connect, temporary on database postgres to my_role;

postgres database

All CAST AIP products will expect the CAST AIP schemas to be members of the postgres database - i.e. all connections to a PostgreSQL instance will connect direct to the postgres database. This database is created by default during the installation of the PostgreSQL instance and should be retained. While you can create an alternative custom database, it is not possible to connect to a custom database using any CAST AIP products.

CAST AIP schemas

Using CAST Server Manager, connect to the 'postgres' database and follow the installation instructions in Install CAST AIP schemas. Please make sure you are using the correct port and connection string when connecting to the PostgreSQL Server:

How to log all statement's plan

Change the following parameters in the configuration file  (..\db_data\postgresql.conf):

auto_explain.log_min_duration = '0s'            # setting this to zero logs all plan
auto_explain.log_nested_statements = on         # log statements executed inside a function
auto_explain.log_verbose = on                   # it's equivalent to the VERBOSE option of EXPLAIN
auto_explain.log_buffers = on                   # it's equivalent to the BUFFERS option of EXPLAIN
auto_explain.log_analyze = on                   # causes EXPLAIN ANALYZE output; when this parameter is on, per-plan-node timing occurs for all statements executed, whether or not they run long enough to actually get logged. This can have an extremely negative impact on performance. Turning off auto_explain.log_timing ameliorates the performance cost, at the price of obtaining less information.
shared_preload_libraries = 'auto_explain'       # change requires restart

Set this parameter to off when only actual row counts, and not exact times, are needed:

auto_explain.log_timing = off                   #  this parameter has no effect unlessauto_explain.log_analyze is enabled