HeartBeat system configuration


  Heartbeat is an application which monitors two nodes using a technique
  with the same name. If the node doesn't send messages ("heartbeats")
  in a certain time, it's considered not active (failed) and the
  services are taken by another node. This process is called failover.

  The hosts that make up this cluster, this coordinated set of machines
  built for a certain goal, are called nodes.


  11..  NNooddeess aanndd SSeerrvviicceess

  This item allows you to configure all the information referring to the
  nodes that make up the cluster and the services provided by those.

  Each node can have resources associated to them. These resources are
  basically virtual IP mumbers that are activated and services to be
  started, when at least one of the nodes in the cluster get started.
  Those IP addresses are disabled when a transition occurs.

  The services are normally standard System V startup scripts, located
  in /etc/rc.d/init.d or /etc/ha.d/resource.d. These scripts, when
  called, receive the "start" or "stop" parameters based on the service
  takeover or its return.  To pass other parameters to the script you
  can use the following sintax: service::parameter.

  It's important to remember that two different nodes can't answer to
  the same IP address simultaneously.




  22..  AAuutthheennttiiccaattiioonn KKeeyyss

  In this item is configured the authentication method, and eventually
  the key, used in the messages sent.

  According to the heartbeat author the SHA1 is considered the best
  method, followed by MD5. The SHA1 and MD5 need that a key be defined.




  33..  DDeevviiccee CCoonnffiigguurraattiioonn

  Here you configure all the devices to be used to send the
  "heartbeats". The devices can be network interfaces or serial ports.
  The serial ports can use a simple protocol or ppp-udp. In the case of
  using the ppp-udp ports, it's also necessary to define an IP address
  to this interface.



  44..  MMiisscc.. CCoonnffiigguurraattiioonnss

  These are the basic configurations:

  * The file name to write debugging messages (optional): Debug messages
  generated by the heartbeat will be recorded in this file.

  * The file name to write other messages (optional): The file where all
  the log messages and warnings will be recorded.


  * Facility to be used with the syslog/logger (optional):

  * Watchdog device (optional): If this node sends no sign to the
  watchdog device for a minute, it will be restarted (reboot).

  * How many seconds between "heartbeats" (optional): This option
  configures the keepalive item, which defines the amount of time in
  seconds on which a heartbeat will be sent to the cluster.

  * Time (s) to declare a node dead (optional): Configures the deadtime,
  the time in seconds without receiving messages from a determined node
  after which it will be declared dead.

  * Number of hops, for ring topologies (optional): (hopfudge) The
  maximum number of hops less the number of nodes in the configuration
  file.

  * Transmission Rate (baud rate) for serial ports (optional): Defines
  the transmition rate of the serial ports used.

  * UDP Port (optional): The UDP Port used for communication between the
  nodes.


  * Nice_failback Controls the system behaviour in case of a failback,
  return after fail, or primary server initialization. If this option is
  enabled, when the primary server - the one which has resources
  associated to its name in the configuration files - is in the
  initialization process, it checks out if the cluster is already
  active. In the case of an affirmative answer and some node is already
  taking care of the resources which correspond to this server, this one
  acts like the secondary. This option is an alternative to the standard
  heartbeat behaviour, on which the primary server *always* takes its
  resources back on restart.