IBM SG24-5131-00 User Manual
158
IBM Certification Study Guide AIX HACMP
node. Because the command is executed remotely, there can be a delay
before the command output is returned.
before the command output is returned.
8.2.3.1 When to Stop Cluster services
You typically stop cluster services in the following situations:
You typically stop cluster services in the following situations:
• Before making any hardware or software changes or other scheduled
node shutdowns or reboots. Failing to do so may cause unintended cluster
events to be triggered on other nodes.
events to be triggered on other nodes.
• Before certain reconfiguration activity. Some changes to the cluster
information stored in the ODM require stopping and restarting the cluster
services on
services on
all nodes for the changes to become active. For example, if
you wish to change the name of the cluster, the name of a node, or the
name of an adapter, you must stop and restart the cluster.
name of an adapter, you must stop and restart the cluster.
8.2.3.2 Types of Cluster Stops
When you stop cluster services, you must also decide how to handle the
resources that were owned by the node you are removing from the cluster.
You have the following options:
When you stop cluster services, you must also decide how to handle the
resources that were owned by the node you are removing from the cluster.
You have the following options:
Graceful
In a graceful stop, the HACMP software shuts
down its applications and releases its resources.
The other nodes do not take over the resources of
the stopped node.
down its applications and releases its resources.
The other nodes do not take over the resources of
the stopped node.
Graceful with Takeover
In a graceful with takeover stop, the HACMP
software shuts down its applications and releases
its resources. The surviving nodes take over these
resources. This is also called
software shuts down its applications and releases
its resources. The surviving nodes take over these
resources. This is also called
intentional failover.
Forced
In a forced stop, the HACMP daemons only are
stopped, without releasing any resources. For
example, the stopped node stays on its service
address if IP Address Takeover has been enabled.
It does not stop its applications, unmount its file
systems or varyoff its shared volume groups. The
other nodes do not take over the resources of the
stopped node. Please note that the forced option
is currently not supported at the Version 4.3 level
in HACMP/ES, only in HACMP Classic.
stopped, without releasing any resources. For
example, the stopped node stays on its service
address if IP Address Takeover has been enabled.
It does not stop its applications, unmount its file
systems or varyoff its shared volume groups. The
other nodes do not take over the resources of the
stopped node. Please note that the forced option
is currently not supported at the Version 4.3 level
in HACMP/ES, only in HACMP Classic.
8.2.3.3 Abnormal Termination of a Cluster Daemon
If the SRC detects that any HACMP daemon has exited abnormally (without
being shut down using the
If the SRC detects that any HACMP daemon has exited abnormally (without
being shut down using the
clstop
command), it executes the
/usr/sbin/cluster/utilities/clexit.rc
script to halt the system. This