Veritas InfoScale™ 7.3.1 Installation Guide - Linux
- Section I. Introduction to Veritas InfoScale
- Section II. Planning and preparation
- System requirements
- Preparing to install
- Setting up the private network
- Setting up shared storage
- Planning the installation setup for SF Oracle RAC and SF Sybase CE systems
- Section III. Installation of Veritas InfoScale
- Installing Veritas InfoScale using the installer
- Installing Veritas InfoScale using response files
- Installing Veritas Infoscale using operating system-specific methods
- Completing the post installation tasks
- Section IV. Uninstallation of Veritas InfoScale
- Uninstalling Veritas InfoScale using the installer
- Uninstalling Veritas InfoScale using response files
- Section V. Installation reference
- Appendix A. Installation scripts
- Appendix B. Tunable files for installation
- Appendix C. Troubleshooting installation issues
Planning your network configuration
The following practices are recommended for a resilient network setup:
Configure the private cluster interconnect over multiple dedicated gigabit Ethernet links. All single point of failures such as network interface cards (NIC), switches, and interconnects should be eliminated.
The NICs used for the private cluster interconnect should have the same characteristics regarding speed, MTU, and full duplex on all nodes. Do not allow the NICs and switch ports to auto-negotiate speed.
Configure non-routable IP addresses for private cluster interconnects.
The default value for LLT peer inactivity timeout is 16 seconds.
For SF Oracle RAC: The value should be set based on service availability requirements and the propagation delay between the cluster nodes in case of campus cluster setup. The LLT peer inactivity timeout value indicates the interval after which Veritas InfoScale on one node declares the other node in the cluster dead, if there is no network communication (heartbeat) from that node.
The default value for the CSS miss-count in case of Veritas InfoScale is 600 seconds. The value of this parameter is much higher than the LLT peer inactivity timeout so that the two clusterwares, VCS and Oracle Clusterware, do not interfere with each other's decisions on which nodes should remain in the cluster in the event of network split-brain. Veritas I/O fencing is allowed to decide on the surviving nodes first, followed by Oracle Clusterware. The CSS miss-count value indicates the amount of time Oracle Clusterware waits before evicting another node from the cluster, when it fails to respond across the interconnect.
For more information, see the Oracle Metalink document: 782148.1