12.5.3.2. Procedure – Configure additional Heartbeat parameters

  1. Configuring timing parameters

    Figure 12.19. Configuring timing parameters

    Set timers for Heartbeat in the Timers section.

    You can define the following timers. Give all the values in seconds.

    keepalive

    The interval in seconds between subsequent Heartbeat signals.

    warning time

    The interval in seconds after which the slave node gives warning of the dead master node.

    dead time

    The interval in seconds after which the node is assumed to be dead.

    initial dead time

    First deadtime after system reboot in seconds.

    Tip

    It may take a while in some machines to have a fully functional network after a reboot. It is recommended to set the dead time twice the warning time, and the initial deadtime twice the dead time. For example, if the warning time is 20 seconds, the optimal dead time is 40 seconds and the optimal initial dead time is 80 seconds.

  2. Configuring other the Heartbeat parameters

    Figure 12.20. Configuring other the Heartbeat parameters

    Set other Heartbeat options in the Misc subwindow.

    You can select log target, watchdog device and decide about service re-acquiry in case of master node recovery.

    log facility

    This option determines the target device for logging.

    watchdog

    This option determines which watchdog device is used, if any. For example, /dev/watchdog.

    Watchdogs can monitor the heartbeat messages of a node and reboot the node if it fails.

    nice fallback

    This option prevents the master node from re-acquiring cluster resources after a failover in case the master node gets functional again. Enabling nice_failback can cause problems in certain situations. Suppose there is a two-node cluster with a master (Node_A) and a slave (Node_B) node. If Node_A fails, Node_B acquires its resources and uses a STONITH device to power off the Node_A. When Node_A recovers, the resources remain on Node_B if nice_failback is enabled. However, if now Node_B seems to fail, Node_A can power off Node_B only if two STONITH devices are intalled to the system (one to power off Node_A, one to power off Node_B).

    Tip

    It is recommended to enable nice_failback, but after a takeover restore the original master-slave node hierarchy at the earliest possible time (for example, during the next maintenance break).