|
|  以 PDF 格式下載這本書 (1391 KB)
scconf(1M)
NAME
| SYNOPSIS
| DESCRIPTION
| OPTIONS
| USAGE
| EXAMPLES
| EXIT STATUS
| ATTRIBUTES
| SEE ALSO
| WARNINGS
| NOTES
NAME
scconf– update the Sun Cluster software configuration
SYNOPSIS
scconf -a [-Hv] [-h node_options] [-A adapter_options] [-B junction_options] [-m cable_options] [-P privatehostname_options] [-q quorum_options] [-D devicegroup_options] [-T authentication_options]
scconf -c [-Hv] [-C cluster_options] [-A adapter_options] [-B junction_options] [-m cable_options] [-P privatehostname_options] [-q quorum_options] [-D devicegroup_options] [-T authentication_options] [-w heartbeat_options]
scconf -r [-Hv] [-h node_options] [-A adapter_options] [-B junction_options] [-m cable_options] [-q quorum_options] [-D devicegroup_options] [-T authentication_options]
scconf -p [-Hv[v]]
scconf [-H]
DESCRIPTION
The scconf command manages the Sun Cluster software
configuration. You can use scconf to add items to the configuration,
to change properties of previously configured items, and to remove items from
the configuration. In each of these three forms of the command, options are
processed in the order in which they are typed on the command line. All updates
associated with each option must complete successfully before the next option
is considered.
The scconf command can only be run from an active
cluster node. As long as the node is active in the cluster, it makes no difference
which node is used to run the command. The results of running the command
are always the same, regardless of the node used.
The -p option of scconf enables you
to print a listing of the current configuration.
All forms of the scconf command accept the -H option. Specifying -H displays help information,
and all other options are ignored and not executed. Help information is also
printed when scconf is invoked without options.
OPTIONS
Basic Options
The following option is common to all forms of the scconf
command:
-
-H
-
If this option is specified on the command line at any position, prints help
information. All other options are ignored and are not executed. Help information
is also printed if scconf is invoked with no options.
The following options modify the basic form and function
of the scconf command. None of these options can be combined
on the same command line.
-
-a
-
Specifies
the add form of the scconf command.
The -a option can be used to add or initialize most of the
items that are used to define the software configuration of a Sun Cluster.
Additional options are used with -a to specify elements (adapter,
junction, or device group options, for example) and their associated properties
to be added. Any number of these additional options can be combined on the
same command line, as long as they are for use with the -a
option.
-
-c
-
Specifies the change form of the scconf command. The -c option is used to change properties of items already configured
as part of the Sun Cluster software configuration. Additional options are
used with -c to specify new or changed properties. Any number
of these additional options can be combined on the same command line, as long
as they are for use with the -c option.
-
-p
-
Specifies the print form of the scconf command. The -p option prints a listing of the current Sun Cluster configuration
elements and their associated properties that you can configure with scconf. This option can be combined with one or more -v
options to print more verbose listings.
-
-r
-
Specifies the remove form of the scconf command. The -r option is used to remove items from the Sun Cluster software configuration.
Additional options are used with -r to specify the items to
delete from the configuration. Any number of these additional options can
be combined on the same command line, as long as they are for use with the -r option.
Additional Options
The following additional options can be combined with one or more of
the previously described basic options. Refer to the SYNOPSIS section to see
the options that can be used with each form of scconf.
The additional options are as follows:
-
-A adapter_options
-
Adds, removes, or changes the properties of a cluster transport
adapter. The node on which the given adapter is hosted need not be active
in the cluster for these operations to succeed. The -A adapter_options for each of the three forms of the command that
accept -A are described here.
-
Use this syntax to specify -A adapter_options for the add form of the command:
–A trtype=type,name=name,node=node[,other_options]
|
-
Use this syntax to specify -A adapter_options for the change form of the
command:
–A name=adaptername,node=node[,state=state] \
[,other_options]
|
-
Use this syntax to specify -A adapter_options for the remove form of the
command:
The -A option supports the following suboptions:
-
trtype=type
-
Specifies the transport type. This suboption must be included
when -A is used with the add form of the
command.
An example of a transport type is dlpi. See sctransp_dlpi(7P).
-
name=adaptername
-
Specifies the name of an adapter on a particular node. This
suboption must be included with each occurrence of the -A
option.
adaptername is constructed from a device name, immediately followed by a physical-unit number (for example, hme0).
-
node=node
-
Specifies the name of an adapter on a particular node. A node suboption is required for each occurrence of the -A option.
The node can be given either as a node name
or node ID.
-
state=state
-
Changes the state of the adapter. You can use this suboption
with the change form of the command. The state can be set to either enabled or disabled.
When an adapter is added to the configuration, its state is always set
to disabled. By default, adding a cable to any of the ports
on an adapter changes the state of both the port and the adapter to enabled. See -m cable_options.
Disabling an adapter also has the effect of disabling all ports associated
with that adapter. However, enabling an adapter does not result in the enabling
of its ports. To enable an adapter port, you must enable the cable to which
the port is connected.
- [other_options]
-
If other options are available for a particular adapter type, they can
be used with -A in the add and change forms of the command.
Refer to the cluster transport adapter man pages (for example, scconf_transp_adap_hme(1M), scconf_transp_adap_eri(1M), and scconf_transp_adap_sci(1M)) for information about special options.
You need solaris.cluster.transport.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-B junction_options
-
Adds, removes, or changes the properties of a cluster transport
junction.
Examples of such devices can include, but are not limited to, Ethernet
hubs, other switches of various types, and rings.
The -B junction_options for
each of the three forms of the command that accept -B are
described here.
-
Use this syntax to specify -B junction_options for the add form of the
command:
–B type=type,name=name[,other_options]
|
-
Use this syntax to specify -B junction_options for the change form of the
command:
–B name=name[,state=state][,other_options]
|
-
Use this syntax to specify -B junction_options for the remove form of the
command:
The -B option supports the following suboptions:
-
type=type
-
Specifies a cluster transport junction type. This suboption
must be included when -B is used with the add
form of the command.
Ethernet hubs and SCI switches are examples of cluster transport junctions
of type switch. The man pages scconf_transp_jct_dolphinswitch(1M) and scconf_transp_jct_etherswitch(1M) contain more information.
-
name=name
-
Specifies the name of a cluster transport junction. A name suboption must be included with each occurrence of the -B option.
name can be up to 256 characters in length.
It is made up of either letters or digits, with the first character being
a letter. Each transport junction name must be unique across the namespace
of the cluster.
-
state=state
-
Changes the state of a cluster transport junction.
This suboption can be used with a -B change
command. state can be set to either enabled
or disabled.
When a junction is added to the configuration, its state is always set
to disabled. By default, adding a cable to any of the ports
on a junction changes the state of both the port and the junction to enabled. See -m cable_options.
Disabling a junction also has the effect of disabling all ports associated
with that junction. However, enabling a junction does not result in the enabling
of its ports. To enable a junction port, you must enable the cable to which
the port is connected.
- [other_options]
-
When other options are available for a particular junction type, they
can be used with -B in the add and change forms of the command. Refer to the cluster transport junction
man pages (for example, scconf_transp_jct_dolphinswitch(1M) and scconf_transp_jct_etherswitch(1M)) for information about special options.
You need solaris.cluster.transport.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-C cluster_options
-
Changes the name of the cluster itself.
This option can only be used with the change form of the
command.
Specify cluster_options for the change form of the command as follows:
-C cluster=clustername cluster=clustername
|
This form of the command changes the name of the cluster to clustername.
-
-D devicegroup_options
-
Adds disk device groups to the configuration, changes or resets
properties of existing device groups, or removes groups from the Sun Cluster
device groups configuration. Other disk device group options (other_options) play a crucial role in adding or changing device
groups and their options. Pay special attention to the man pages for the type-dependent
disk device group options (for example, scconf_dg_vxvm(1M), scconf_dg_sds(1M), scconf_dg_svm(1M),
and scconf_dg_rawdisk(1M)) when configuring
any device group. Not all device group types support all three forms of the -D option. For example, sds device groups can normally
only be used with the change form of the command to change
certain attributes, such as the ordering of the node preference list.
The add form of the command can be used to either
create device groups or to add nodes to existing device groups. For some device
group types, the add form can also be used to add devices
to a group. The change form of the command registers updates
to change certain attributes associated with a group. The remove form of the command is used to either remove an entire device group
or one or more of a group's components.
The -D devicegroup_options
for each of the three forms of the scconf command that
accept -D are as follows:
Add:
-D type=type,name=name,nodelist=node[:node]…
[,preferenced={true | false}]
[,numsecondaries=integer]
[,failback={enabled | disabled}][,other_options]
|
Change:
-D name=name[,nodelist=node[:node]…]
[,preferenced={true | false}]
[,numsecondaries=integer]
[,failback={enabled | disabled}][,other_options]
|
Remove:
-D name=name,nodelist=node[:node]…
|
The -D option supports the following suboptions:
-
type=type
-
Must be used with the add form of the command
to indicate the type of disk device group to create (for example, vxvm or rawdisk).
-
name=name
-
Is the name of the disk device group
and must be supplied with all three forms of the command.
-
nodelist=node[:node]…
-
Is a list of potential primary
nodes that is required for some disk device group types when adding a group
to the cluster. Refer to the man pages for the type-dependent disk device
group for more information.
With the add form of the command, the nodelist is, by default, an ordered list indicating the preferred order
in which nodes should attempt to take over as the primary node for a disk
device group. However, if the preferenced suboption is
set to false (see the next subsection), the first node
to access a device in the group automatically becomes the primary node for
that group. The preferenced suboption cannot be used when
adding nodes to an existing device group. However, the preferenced suboption can be used when you create the group for the first time,
or with the change form of the command.
To change the primary node order preference, you must specify the complete
list of cluster nodes in the nodelist in the order that
you prefer. You must also set the preferenced suboption
to true.
When used with the remove form of the command, the nodelist suboption is used to remove the indicated nodes from the
device group. Only by not providing a nodelist can the
entire device group be removed. Simply removing all of the nodes from a device
group does not necessarily remove that group.
- [preferenced={true | false}]
-
Indicates the status of the preferred
order of potential primary nodes for a disk device group. As long as the preferenced suboption is not set to false, node
lists for newly created device groups indicate a preferred order in which
nodes attempt to take over as the primary node for a disk device group.
If the preferenced suboption is not specified with
an add that is used to create a device group, it is, by
default, false. However, if the preferenced suboption is
not specified with a change, it is, by default, set to true when nodelist is given.
The preferenced suboption cannot be used with an add that is used to add nodes to an established device group. In
this case, the established node preference list setting is used.
- [numsecondaries=integer]
-
Enables you to dynamically change the
desired number of secondary nodes for a device group. A device group is an
HA service that requires one node to act as a primary node and one or more
nodes to act as secondary nodes. The secondary nodes of a device group are
able to take over and act as the primary node if the current primary node
fails.
This integer value should be greater than 0 but less than the total
number of nodes in the specified group. The default is 1.
A system administrator can use the numsecondaries
suboption to change the number of secondary nodes for a device group while
maintaining a given level of availability. If a node in a device group is
removed from the secondary nodes list, it is not able to take over and act
as a primary node until it is converted back to a secondary node. Before making
a change to the number of secondary nodes, you need to assess the impact on
the secondary global file system.
The numsecondaries suboption only applies to nodes
in a device group that are currently in cluster mode and can be used together
with the node's preferenced suboption. If a device's preferenced suboption is enabled, the nodes that are least preferred
are removed from the secondary nodes list first. If no node in a device group
is flagged as preferred, the cluster randomly picks the node to remove.
When a device group's actual number of secondary nodes drops to less
that the desired level due to node failures, nodes that were removed from
the secondary nodes list are added back to the secondary list of nodes if
they are currently in a cluster, belong to the device group, and are not currently
a primary or a secondary node. The conversion starts with the node in the
device group with the highest preference until the number of desired secondary
nodes is matched.
If a node in the device group has a higher preference than an existing
secondary node and joins the cluster, the node with the least preference is
removed from the secondary nodes list and is replaced by the newly added node.
This replacement only occurs when there are more actual secondary nodes than
the desired level.
To set the desired number of secondary nodes to the system default (without
having to know the default value), issue one of these commands:
# scconf -aD type=vxvm,name=foo, \
nodelist=node1:node2,numsecondaries=
|
or
# scconf -cD name=foo,numsecondaries=
|
The numsecondaries suboption can only be used with
the -a option when a device group is created. The numsecondaries suboption cannot be used with the -a
option to add a host to an existing device group.
- [failback={enabled | disabled}]
-
Enables or disables the failback behavior of a disk device group with either the add or the change form of the command.
Specifies the behavior of the system should a disk device group primary
node leave the cluster membership and later return.
When the node leaves the cluster membership, the disk device group fails
over to the secondary node. When the failed node rejoins the cluster membership,
the disk device group can either continue to be mastered by the secondary
node, or fail back to the original primary node.
If failback is enabled, the disk
device group becomes mastered by the original primary node. If failback is disabled, the disk device group continues
to be mastered by the secondary node.
By default, failback is disabled.
- [other_options]
-
You can use other disk device group type-dependent options with either
the add or change form of the command.
Refer to the appropriate man pages for more information (for example, scconf_dg_vxvm(1M), scconf_dg_sds(1M), scconf_dg_svm(1M), and scconf_dg_rawdisk(1M)).
You need solaris.cluster.device.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-h node_options
-
Adds or removes a node from the cluster configuration database.
When used with the add form of scconf,
both the new name and an internally generated node ID are added to the cluster
configuration database. In addition, the new node is given a disk reservation
key and a quorum vote count of zero. The name that is assigned to access the
node over the cluster interconnect is initialized to clusternodenodeid-priv. See the -p option to
learn more about printing configuration elements and their associated properties.
scconf cannot be used by itself to add a new node
to the cluster. You can only use scconf to update the configuration
database itself. scconf does not copy the configuration
database onto the new node or create the necessary node identifier on the
new node. To add a node to a cluster, use scinstall(1M).
When used with the remove form of scconf, all references to the node, including the last transport cable,
all resource group references, and all device group references must be removed
before scconf can be used to completely remove the node
from the cluster configuration.
The node to be removed must not be configured for any quorum devices.
In addition, you cannot remove a node from a three-node cluster unless there
is at least one shared quorum device configured.
The system administration procedures in the Sun Cluster documentation
describe how to remove a cluster node in more detail.
You must specify the node=node
suboption with any occurrence of the -h option. For the add form of the command, the given node
must be a node name.
Use this syntax to specify the -h node_options for the add form of the command:
For the remove form of the command, the node can be given either as a node name or node ID. Use this
syntax to specify the -h node_options
for the remove form of the command:
You need solaris.cluster.node.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-m cable_options
-
Helps to establish the cluster interconnect topology. This
option helps by configuring the cables that are connecting the various ports
that are found on the cluster transport adapters and junctions. Each new cable
typically maps a connection either between two cluster transport adapters
or between an adapter and a port on a transport junction. The -m cable_options for each of the forms of the command that accept -m are as follows:
-
Use this syntax to specify the -m cable_options for the add form of the command:
–m endpoint=[node:]name[@port],
endpoint=[node:]name[@port][,noenable]
|
-
Use this syntax to specify the -m cable_options for the change form of the
command:
–m endpoint=[node:]name[@port],state=state
|
-
Use this syntax to specify the -m cable_options for the remove form of the
command:
–m endpoint=[node:]name[@port]
|
The -m option supports the following suboptions:
-
endpoint=[node:]name[@port]
-
Must be included with each occurrence of the -m option. For the add form of the command, two endpoint options must be specified. The name
component of the option argument is used to specify the name of either a cluster
transport adapter or cluster transport junction at one of the endpoints of
a cable. If a node component is given, the name is the name of a cluster transport adapter. Otherwise,
the name is the name of a cluster transport junction.
If a port component is not given, an attempt
is made to assume a default port name. The default port for an adapter is
always 0. The default port name for a junction endpoint
is equal to the node ID of the node attached to the other end of the cable.
Refer to the cluster transport adapter and cluster transport junction man
pages for more information about port assignments
and other requirements (for example,scconf_transp_adap_hme(1M), scconf_transp_adap_eri(1M), scconf_transp_adap_sci(1M), scconf_transp_jct_etherswitch(1M), and scconf_transp_jct_dolphinswitch(1M)). Before a cable can be added, the adapters
and junctions at each of the two endpoints of the cable must already be configured
(see -A and -B).
-
state=state
-
Changes the state of a cable and the two endpoints
to which it is connected. When a cable is enabled, the cable, its two ports,
and the adapters or junctions associated with those two ports are all enable. However, when a cable is disabled, only
the cable and its two ports are disabled. The state of the adapters or junctions
associated with the two ports remains unchanged. By default, the state of
a cable, and its endpoints, is always set to enabled at
the time that the cable is added to the configuration. But, to add a cable
in the disabled state, use noenable
as part of an add.
-
noenable
-
Can
be used when adding a cable to the configuration. By default, when you add
a cable, the state of the cable, the two ports to which it is connected, and
the adapters or junctions on which the ports are found, are set to enable. But, if noenable is specified when you
add a cable, the cable and its two endpoints are added in the disabled state.
The state of the adapters or junctions on which the ports are found remains
unchanged.
You need solaris.cluster.transport.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-P privatehostname_options
-
When used with either the add or change form of the command, specifies a host
name alias to use for IP access of a given node over the private cluster interconnect,
or transport. If not otherwise assigned, or if reset, the default private
host name is clusternodenodeid-priv.
Private host names should never be stored in the hosts(4) database. A special nsswitch facility
(see nsswitch.conf(4))
performs all host name lookups for private host names.
Both the add and change forms
of scconf behave identically in relation to the -P option. The -P privatehostname_options for each of the two forms of the command that accept -P are as follows:
Add:
–P node=node[,privatehostname=hostalias]
|
Change:
–P node=node[,privatehostname=hostalias]
|
The -P option supports the following suboptions:
-
node=node
-
Provides the name or ID of the node to be assigned
the private host name, or host alias, supplied with the privatehostname suboption.
- [privatehostname=hostalias]
-
Supplies the host alias to be used for
accessing a node over the private cluster interconnect, or transport. If no privatehostname suboption is specified, the private host name for
the given node is reset to the default.
You need solaris.cluster.transport.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-q quorum_options
-
Manages shared cluster quorum devices and various cluster
quorum properties. The add and remove
forms of the command are used to add and remove shared quorum devices to or
from the configuration. The change form of the command
is used for changing various cluster quorum configuration properties or states.
The -q quorum_options available
for each of the three forms of the command that can be used to change the
cluster quorum configuration are as follows:
Add:
–q globaldev=devicename[,node=node,node=node[,...]]
|
Change:
–q node=node,{maintstate | reset}
–q globaldev=devicename,{maintstate | reset}
–q reset
–q installmode
|
Remove:
When scconf is interrupted or fails while performing
quorum-related operations, quorum configuration information can become inconsistent
in the cluster configuration database. If this occurs, either run the same scconf command again or run it with the reset
option to reset the quorum information.
The -q option supports the following suboptions:
-
globaldev=devicename
-
Specifies the name of a global disk
device to use when adding or removing a shared quorum device to or from the
cluster. This suboption can also be used with the change
form of the command to change the state of a quorum device.
Each quorum device must be connected, or ported, to at least two nodes
in the cluster. It is not possible to use a non-shared disk as a quorum device.
With the add form of the command, if a globaldev is specified without a node list, the
quorum device is added with a port defined for every node to which the device
is attached. But, if a node list is given, at least two
nodes must be provided. And, each node in the list must be ported to the device.
The change form of scconf can
be used with -q globaldev to either put
the device into a maintenance state or to reset the device's quorum configuration
to the default. While in maintenance state, the device takes on a vote count
of zero and, so, does not participate in forming quorum. When reset to the
default, the vote count for the device is changed to N-1, where N is the number
of nodes with nonzero vote counts that have ports to the device.
-
node=node
-
When used with the add form of the
command, selects the nodes that should be configured with ports to the shared
quorum device being added. This suboption can also be used with the change form of the command to change the quorum state of a node.
When the node suboption is used with the change form of the quorum update command, it is used to either place
a node into maintenance state or to reset the node's quorum configuration
to the default.
You must shut down a node before you can put it into maintenance state. scconf returns an error if you attempt to put a cluster member into
maintenance state.
While in maintenance state, the node takes on a vote count of zero and,
so, does not participate in quorum formation. In addition, any shared quorum
devices configured with ports to the node have their vote counts adjusted
down by one to reflect the new state of the node. When the node is reset to
the default, its vote count is reset to 1 and the shared quorum device vote
counts are re-adjusted back up. Unless the cluster is in installmode, the quorum configuration for each node is automatically reset
at boot time.
A node can be specified either as a node
name or node ID.
- {maintstate}
-
When used as a flag with the change form of the command,
for either the globaldev or node suboptions,
puts a shared quorum device or node into a quorum maintenance state. When
in maintenance state, a shared device or node no longer participates in quorum
formation. This feature can be useful when a node or device must be shut down
for an extended period of maintenance. Once a node boots back into the cluster,
under usual circumstances, it removes itself from maintenance mode.
It is not legal to specify both maintstate and reset with the same -q option.
- {reset}
-
When
used as a flag with the change form of the command, resets
the configured quorum vote count of a shared quorum device or node. This option
can be combined with either the globaldev or node suboptions, or it can be its own suboption.
If used by itself, the entire quorum configuration is reset to the default
vote count settings. In addition, if installmode is set,
it is cleared by a global quorum configuration reset. installmode cannot be reset on a two-node cluster unless at least one shared
quorum device has been successfully configured.
-
installmode
-
Forces the cluster back into installmode. While in installmode, nodes do not attempt to reset their quorum configurations
at boot time. Also, while in this mode, many administrative functions are
blocked. When a cluster is first installed, it is set up with installmode set. Once all of the nodes have joined the cluster for the first
time, and shared quorum devices have been added to the configuration, issue scconf -c -q reset to reset the vote counts to their
default values and to clear the installmode setting.
You need solaris.cluster.quorum.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-T authentication_options
-
Establishes authentication policies for
nodes that are attempting to add themselves to the cluster configuration.
Specifically, when a machine requests that it be added to the cluster as a
cluster node (see scinstall(1M)),
a check is made to determine whether or not the node has permission to join.
If the node has permission, the joining node is authenticated. By default,
any machine is allowed to add itself to the cluster.
The -T authentication_options
for each of the three forms of the command that accept -T
are as follows:
Add:
–T node=nodename[,...][,authtype=authtype]
|
Change:
Remove:
–T {node=nodename[,...] | all}
|
The -T option supports the following suboptions:
-
node=nodename
-
Adds or removes host names from the
list of nodes that are able to install and configure themselves as nodes in
the cluster. At least one node suboption is required for
the add form of the command and is optional for remove. If the authentication list is empty, any host can request
that it be added to the cluster configuration. However, if the list has at
least one name in it, all such requests are authenticated using the authentication
list.
Illegal nodenames are accepted, including
the node name of dot (.). The dot character is special
in that if a nodename of . is
added to the authentication list, all other names are removed. This feature
prevents a host from attempting to install and configure itself in the cluster.
-
all
-
You can clear
the list of all node names by specifying scconf -r -T all. A cleared authentication list means that any node can attempt
to install and configure itself in the cluster.
-
authtype=authtype
-
Is used with either the add or change form of the command.
The only currently supported authentication types (authtype) are des and sys (or unix). The default authentication type is sys,
which provides the least amount of secure authentication.
When des, or Diffie-Hellman, authentication is used,
entries should be added to the publickey(4)
database for each cluster node to be added before actually running scinstall(1M)
to add the node.
You need solaris.cluster.node.modify RBAC authorization
to use this command option with -a, -c, or -r. See rbac(5).
-
-v
-
When used with
the -p option, requests a more verbose, or detailed, listing
of the cluster configuration. If used with other options, additional information
might be printed when an error is encountered.
You need solaris.cluster.device.read, solaris.cluster.transport.read, solaris.cluster.resource.read, solaris.cluster.node.read, solaris.cluster.quorum.read, and solaris.cluster.system.read RBAC authorizations to use this command
option with -p. See rbac(5).
-
-w heartbeat_options
-
Changes the global heartbeat parameters of a cluster, which
effectively changes the heartbeat parameters across all the adapters of the
cluster.
Sun Cluster relies on heartbeats over the private interconnect to detect
communication failures among cluster nodes. Reducing the heartbeat timeout
enables Sun Cluster to detect failures more quickly, as the time that is required
to detect failures decreases when you decrease the values of heartbeat timeout.
Thus, Sun Cluster recovers more quickly from failures, consequently increasing
the availability of your cluster.
The -w option supports the following suboptions:
-
heartbeat_quantum=quantum_milliseconds
-
Defines how often to send heartbeats.
Sun Cluster uses a 1 second (1,000 milliseconds) heartbeat quantum by default.
Specify a value between 100 milliseconds and 10,000 milliseconds.
-
heartbeat_timeout=timeout_milliseconds
-
The time interval after which,
if no heartbeats are received from the peer nodes, the corresponding path
is declared as down. Sun Cluster uses a 10 second (10,000 millisecond) heartbeat
timeout by default. Specify a value between 2,500 milliseconds and 60,000
milliseconds.
Note –
Even under ideal conditions, when you reduce the values of heartbeat
parameters with -w, there is always a risk that spurious path
timeouts and node panics might occur. Always test and thoroughly qualify the
lower values of heartbeat parameters under relevant workload conditions before
actually implementing them in your cluster.
USAGE
With the -w option, you can change only one heartbeat
suboption at a time. When decreasing the values of heartbeat parameters, change heartbeat_quantum first, followed by heartbeat_timeout. When increasing the values of heartbeat parameters, change heartbeat_timeout first, followed by heartbeat_quantum.
Note –
The value you specify for heartbeat_timeout
must always be greater than or equal to five times the value you specify for heartbeat_quantum (heartbeat_timeout >= (5*heartbeat_quantum)).
You need solaris.cluster.system.modify RBAC authorization
to use -w. See rbac(5).
If you change heartbeat parameters with -w, and you
later choose to back out and go back to a previous version of Sun Cluster
that does not support -w, you must first reset all heartbeat
parameters to the original, default values that are supported in the previous
version.
EXAMPLES
Example 1 Decreasing the Heartbeat
The following example shows how to decrease the heartbeat quantum to
100 milliseconds from the Sun Cluster default of 1,000 milliseconds. This
example also shows how to decrease the heartbeat timeout to 2500 milliseconds
from the Sun Cluster default of 10,000 milliseconds.
phys-schost-1# scconf -c -w heartbeat_quantum=100
phys-schost-1# scconf -c -w heartbeat_timeout=2500
|
Because heartbeat_timeout must always be greater
than or equal to five times heartbeat_quantum, you need
to set heartbeat_quantum first. Otherwise, the requirement
is not met. In other words, if heartbeat_quantum is currently
set to the default 1,000 milliseconds, and if you were to set heartbeat_timeout to 2500 milliseconds, heartbeat_timeout would
be less than five times heartbeat_quantum.
The scconf command would consequently fail.
Once heartbeat_quantum is set to the correct value
however, the requirement is maintained, and you can then set heartbeat_timeout to the decreased value.
Example 2 Increasing the Heartbeat
The following example shows how to increase the heartbeat timeout and
heartbeat quantum to Sun Cluster default values from the values to which you
set these parameters in the previous example.
phys-schost-1# scconf -c -w heartbeat_timeout=10000
phys-schost-1# scconf -c -w heartbeat_quantum=1000
|
You set heartbeat_timeout first to maintain the requirement
that heartbeat_timeout always be greater than or equal
to five times heartbeat_quantum. Once heartbeat_timeout is set to the value you want, you can then set heartbeat_quantum to the new, increased value.
Example 3 Typical Postinstallation Setup Operations
The following commands provide an example of a typical set of postinstallation
setup operations that might be performed on a new two-node cluster. These
commands add a shared quorum device to the cluster, clear installmode, configure a second set of cluster transport connections, and secure
the cluster against other machines that might attempt to add themselves to
the cluster:
phys-red# scconf -a -q globaldev=d0
phys-red# scconf -c -q reset
phys-red# scconf -a \
-A trtype=dlpi,name=hme1,node=phys-red \
-A trtype=dlpi,name=hme1,node=phys-green \
-m endpoint=phys-red:hme1,endpoint=phys-green:hme1
phys-red# scconf -a -T node=.
|
EXIT STATUS
ATTRIBUTES
SEE ALSO
scconf_dg_rawdisk(1M), scconf_dg_sds(1M), scconf_dg_svm(1M), scconf_dg_vxvm(1M), scconf_transp_adap_bge(1M), scconf_transp_adap_ce(1M), scconf_transp_adap_e1000g(1M), scconf_transp_adap_eri(1M), scconf_transp_adap_ge(1M), scconf_transp_adap_hme(1M), scconf_transp_adap_qfe(1M), scconf_transp_adap_sci(1M), scconf_transp_adap_wrsm(1M), scconf_transp_jct_dolphinswitch(1M), scconf_transp_jct_etherswitch(1M), hosts(4), nsswitch.conf(4), publickey(4), attributes(5), sctransp_dlpi(7P)
WARNINGS
Use the -w option only when all
nodes in a cluster are up. Do not use -w when any node in
a cluster is down. Nodes might hang or panic as a result.
Clusters that contain one or more single-CPU nodes, or that contain
more than eight nodes, are more likely to experience timeouts and node panics
when the clusters run with low heartbeat parameter values.
Note –
Even under ideal conditions, when you reduce the values of heartbeat
parameters with -w, there is always a risk that spurious path
timeouts and node panics might occur. Always test and thoroughly qualify the
lower values of heartbeat parameters under relevant workload conditions before
actually implementing them in your cluster.
NOTES
You should either back up the root file system on every node after changing
the configuration with scconf, or keep a log of all changes.
If you need to recover configuration changes between normal system backups,
use the log to return to the most recent configuration.
Option lists specified with the scconf command are
always executed in the order that you specify them on the command line. But,
whenever possible, certain transport options (-A, -B, and -m) are processed by scconf
as a single transaction against the cluster configuration database. Try to
group all related options of this type together on a single command line to
reduce overhead to the cluster.
The -w option works only in 3.1 8/04 and later
versions of Sun Cluster that run on Solaris 8 Update 7 and later versions
of Solaris.
Sun Cluster 3.1 Last Revised 7 Jun 2004
NAME
| SYNOPSIS
| DESCRIPTION
| OPTIONS
| USAGE
| EXAMPLES
| EXIT STATUS
| ATTRIBUTES
| SEE ALSO
| WARNINGS
| NOTES
|
