On the node holding the master role, a logical address is assigned to
an interface that is used to connect to the external network. This address
is called the floating external address. The Node State
Manager (NSM) uses the Cluster Membership Manager (CMM) notifications to determine
when a node is promoted to or demoted from being the master node. When the
NSM is notified that a node has been promoted to master node, it executes
a script to configure a floating external address for one of the node's external
interfaces. When the NSM is notified that a node has been demoted from the
master node, it executes a script to deconfigure the floating external address.
The NSM can be used for tasks other than address management but this
is not its intended function. If you configure the NSM for purposes other
than to monitor the floating external address of the master node, you must
consider the effect of this mechanism on cluster integrity.
Launch the nhnsmd manually as a superuser:
# ./nhnsmd -u URL
The nhnsmd daemon is started at system boot time
after the nhcmmd daemon. The nhnsmd
daemon registers to receive the following notifications at cluster startup:
-
CMM_MASTER_ELECTED
-
CMM_MASTER_DEMOTED
-
CMM_VICEMASTER_ELECTED
-
CMM_VICEMASTER_DEMOTED
The nhnsmd daemon executes a response to notifications
in the order in which it receives them. It does not act upon a notification
that does not pertain to the current state of the node for which the notification
is received.
The nhnsmd daemon maintains persistent state across
failures so that when restarted by the Daemon Monitor it can determine whether
it has missed any notifications and can take appropriate action. This persistent
state is not maintained across a node reboot.
The scripts executed by the nhnsmd daemon must be
executable shell scripts.
The first argument of the scripts is the action
parameter. The action parameter can have two values, enter state and leave
state. You can use the same script for both actions. For an enter
state action, the script passes the string enter
as the first argument. For a leave state action, the
script is passed the string leave as the first argument.
The second argument of the scripts is the node role
parameter. The node role parameter is passed as a lower-case
character string. The node role parameter can have
two values, master and vice-master.
The scripts executed by the nhnsmd daemon should
not perform actions that change the startup behavior of the node.
For any notification received, the nhnsmd daemon executes
a script that invokes a node to leave an existing state and then a script
that invokes that node to enter a new state.
The script provided with Foundation Services handles address failover if the external
master address is present on a separate external interface or on a logical
interface on the cluster network.
Scripts used by the nhnsmd daemon run as asynchronous
processes and do not take account of any changes in the cluster state. When
writing your own action scripts for the nhnsmd daemon do
not write scripts that will take a long time to execute or that depend on
cluster behavior. Such scripts should not be used as a way of controlling
applications or as a replacement for a management framework.
-
nhfs.conf
-
Configuration and addressing information for the different Foundation Services. The
URL for this file could be file:///etc/opt/SUNWcgha/nhfs.conf
-
/opt/SUNWcgha/actions/master
-
This directory contains scripts for transitions to and from
the master state.
-
/opt/SUNWcgha/actions/vicemaster
-
This directory contains scripts for transitions to and from
the vice-master state.
The script names have the Ennxxxxxx or Lnnxxxxxx
form, where E denotes an script to enter a state and L denotes a script to leave a state, nn
is a two-digit numeric code, and xxxxxx is an arbitrary
string of characters. Any files that do not use this naming scheme will be
ignored by nhnsmd.