SharePlex 12.0 - Installation and Setup Guide

Confirm that the platform is supported.

Review the SharePlex Release Notes to make certain your operating system is supported.

Allocate at least 4 GB of memory for SharePlex processes.

Plan for per-process memory up to 256 MB. This recommendation enables the Post and Read processes to allocate larger sets of memory when necessary.

Set the number of semaphores per process.

Semaphores help ensure the stability of the SharePlex processes. The required SharePlex settings depend on the platform, as follows:

Red Hat Linux:

• semmni*: 70
• semmns*: 255
• semmnu: 255
• semmsl: 128
• semopm: 64
• semume: 255
• shmmax: 60 MB
• shmmin: 1MB
• shmmni: 100

• shmseg: 26

*These are additive. Add the database minimum values to the SharePlex minimum values to determine the correct setting.

An alternative is to set the value to the number of queues you will be using plus 2.

Set the ulimit (number of system file descriptors) to as close to 1024 as possible.

The ulimit can be set either as a system hard limit or a session-based soft limit, as follows:

• Set a hard limit: (Recommended) A root user and system restart are required to change the hard limit, but the value remains fixed at the correct level to support SharePlex. Consult your System Administrator for assistance.
• Set a soft limit: A soft limit setting stays in effect only for the duration of the sp_cop session for which it was set, and then it reverts back to a default value that may be lower than the hard limit and too low for SharePlex.

Set soft and hard limits for the SharePlex user

Set a soft limit and a hard limit for nproc and nofile for the SharePlex O/S user in the /etc/security/limits.conf file.

• shareplex O/S user soft nproc 2048
• shareplex O/S user hard nproc 16384
• shareplex O/S user soft nofile 1024
• shareplex O/S user hard nofile 65536

As an alternative, you can simply use the setting for the PostgreSQL O/S user.

Set core file parameters.

• Set the system core dump block size as large as system resources can accommodate, at minimum 1.5 million blocks. The default is usually 0. Core files help Quest support representatives resolve SharePlex support cases. Higher size settings ensure that enough data is captured to be useful.
• Set the core file output location to the dump sub-directory of the SharePlex variable-data directory.
• Set the naming convention of core files to either core or core.pid.

Note: SharePlex renames all core files named core to core.pid, except for those generated by sp_cop.

If these requirements are not met, the SharePlex event log might report that a core file was not generated, even though a file exists.

Install the ksh shell.

Install the ksh shell before you install SharePlex. The SharePlex monitoring scripts and other features required this shell.

A version of ksh called pdksh is included with the Red Hat Linux builds. Refer to the Red Hat Linux documentation for more information.

Install on multiple EBS volumes

• Install the database and SharePlex on Amazon Elastic Block Storage (EBS). An EBS volume is persistent storage, whereas the default Amazon storage is non-persistent and data is lost when the instance shuts down. EBS volumes also provide better performance. Minimum size for a volume is 1GB.
• To optimize disk performance, create multiple EBS volumes and combine them by using software RAID. According to benchmarks, the optimal number of EBS volumes is 8.