PRIMECLUSTER Global Link Services Configuration and Administration Guide: Redundant Line Control Function 4.1 (for Linux)

2.4.1 General

Notes on setting a configuration:

• The minimum and maximum number of virtual and logical virtual interface can be defined is 1 to 64 in total.
• The number of physical interfaces can be used for redundancy on a single virtual interface is within 1 to 8 for Fast switching mode. For NIC switching mode, the range is within 1 to 2.
• The number of logical virtual interfaces that can be defined to a single logical virtual interface is within 1 to 63.
• To use all host names and IP addresses used in a Redundant line Control Function , they must be defined in /etc/hosts files of the local system
• The system automatically determines the length of MTU for an interface. Nonetheless, it is possible to change the length of MTU using user command execution function. For changing MTU length, refer to "3.6.9 Setting user command execution function". Note that the length of MTU cannot be modified in NIC switching mode.

Notes on the operation:

• It is not possible to use a multicast IP address in a Redundant Line Control Function.
• Redundant control line function does not support IPv6-IPv4 tunneling interface(sitX).
• Redundant control line function is comprised with a function monitors conditions of interfaces. If a user manually switches a physical interface up/down, this function overrides user operation and restores these interfaces to the state where they were initially running.
• Do not operate physical interfaces that a virtual interface bundles with an ifconfig command.

Notes on upper applications:

• When using TCP in a working application, the data lost when an error occurred in a transfer route is guaranteed by resending from TCP and reaches the other system in the end. Therefore, TCP connection is not disconnected and there is no error in communication. However, it is necessary to set a timer value longer than the time to finish disconnecting/switching a transfer route when an application monitors a response by such as a timer. When TCP connection is disconnected by the reason such as not possible to change a timer value, reestablish the TCP connection and recover the communication.
• The data lost at the time of an error in a transfer route is not guaranteed when a working application uses the UDP. It is necessary to execute a recovery process such as sending the data by the application itself.
• It is not possible to use DHCP (a server function and a client function) as the upper application in a Redundant Line Control Function.
• When using NTP as an upper application, it is necessary to activate an IP address that a Redundant Line Control Function controls before activating an NTP daemon. No special operation is required when activating a system because a Redundant Line Control Function is activated before an NTP daemon. However, when manually activated an IP address with an operation command or when running cluster operation, reactivate an NTP daemon after an IP address is activated. In addition, when using NTP on GLS, a NTP daemon has to be defined to be able to communicate using a logical IP address. In this case, please add -L option to a configuration file (/etc/sysconfig/ntpd) of NTP daemon.

# Drop root to id 'ntp:ntp' by default. Requires kernel >= 2.2.18.
OPTIONS="-L -U ntp -p /var/run/ntpd.pid"