vxclustadm(1M) VxVM 3.5 vxclustadm(1M)
1 Jun 2002
NAME [Toc] [Back]
vxclustadm - start, stop, and reconfigure a cluster
SYNOPSIS [Toc] [Back]
vxclustadm abortnode
vxclustadm [-s seqno] membership [node_ID ...]
vxclustadm [-v] nodestate
vxclustadm -m {gab|nocm|vcs} [-t {gab|udp}] startnode
vxclustadm stopnode
DESCRIPTION [Toc] [Back]
The vxclustadm utility activates and deactivates cluster functionality
on a node in a cluster. vxclustadm provides an interface to the
clustering functionality of VERITAS Volume Manager (VxVM) in the
absence of a cluster monitor.
Caution: If there is no cluster monitor or if GAB is used as the
cluster monitor, ensure that you completely understand the
functionality of this command. Otherwise, cluster reconfiguration
problems may occur.
KEYWORDS [Toc] [Back]
abortnode Stops clustering activity on a node and abandons all
uncompleted I/O on shared volumes. This command is for
emergency shutdown.
membership
Provides cluster membership information to the VxVM kernel
in the absence of a cluster monitor. This argument must be
issued concurrently on all nodes belonging to the cluster
and on all nodes that you want to bring into the cluster.
Issue membership on all nodes every time cluster membership
has been changed (by startnode, stopnode, or abortnode being
run on a node).
Note: The membership command is valid only for clusters that
are initialized with the nocm argument.
Caution: Ensure that you completely understand the
functionality of the membership command. Otherwise, cluster
reconfiguration problems may occur.
Cluster reconfiguration takes place whenever a node joins or
leaves the cluster. The cluster reconfiguration sequence
number uniquely identifies a new cluster configuration every
time the configuration changes. Use the -s option to
specify the new sequence number.
- 1 - Formatted: January 24, 2005
vxclustadm(1M) VxVM 3.5 vxclustadm(1M)
1 Jun 2002
nodestate Determines the state of a node in the cluster. The node
state is returned as a string to the standard output.
Possible states are:
cluster member
The node is a member of the cluster. All shared
volumes in the cluster are accessible.
joining The node is in the process of joining a cluster.
It has been initialized but is not yet completely
in the cluster. The node goes into this state
after vxclustadm is executed with the startnode
option. For clusters initialized with the nocm
argument, the node remains in this state until
cluster membership is enabled by running
vxclustadm with the membership option. The node
enters the cluster member state after the joining
protocol completes.
out of cluster
The node has not yet joined the cluster.
For debugging purposes the -v option can be specified to
display the node ID, master ID, neighbor ID and current
state.
startnode The startnode option initiates cluster functionality on a
node using the information supplied in the cluster
configuration file. This is the first command that must be
issued on a node to bring it into the cluster. Use the -m
option to specify the cluster monitor.
You must also specify one of the following cluster monitor
types, each of which requires the existence of a cluster
configuration file:
gab The cluster uses GAB (group membership and atomic
broadcast) as the cluster monitor; it is not
running in a VERITAS Cluster Server (VCS)
environment. The cluster configuration file is
/etc/vx/cvmtab.
nocm The cluster does not have a cluster monitor. The
cluster configuration file is /etc/vx/cvmtab.
vcs The cluster is running in a VCS environment. The
cluster configuration file is
/etc/VRTSvcs/conf/config/main.cf.
startnode passes the information in the cluster
configuration file to the VxVM kernel. In response to this
- 2 - Formatted: January 24, 2005
vxclustadm(1M) VxVM 3.5 vxclustadm(1M)
1 Jun 2002
command, the kernel and the VxVM configuration daemon,
vxconfigd, perform the initialization.
Caution: Do not use the nocm and gab cluster monitor types
with the startnode command unless you completely understand
their purpose.
The clustering functionality of VxVM can use either UDP or
GAB as its transport agent for messaging. The default
behavior is to use UDP as the transport agent. The -t
option may be used to specify the protocol to be used for
messaging. This option takes one of the following
arguments:
gab Use GAB as the transport agent for messaging in
addition to using GAB as a cluster monitor. If
you try to use GAB as a transport agent with a
cluster monitor other than GAB (or outside the VCS
environment), the kernel issues a warning message
and changes the transport agent to UDP.
udp Use UDP as the transport agent for messaging
(default).
Both messaging transport agents (GAB and UDP) require that
the /etc/vx/cvmtab file exist. However, the port numbers
and local network names or IP addresses may be omitted (see
the EXAMPLES section below).
In VCS environments, using GAB as the transport agent for
messaging means that you do not need to configure IP
addresses for the endpoints of the cluster private network.
stopnode Stops cluster functionality on a node, and waits for all
outstanding I/O to complete and for all applications to
close shared volumes or devices.
EXIT CODES [Toc] [Back]
vxclustadm returns the following exit values on error:
101 Out of cluster.
102 Joining the cluster.
103 Cluster member.
FILES [Toc] [Back]
For a cluster that is operating without a cluster monitor, or that is
using GAB as the cluster monitor outside a VCS environment, and which
is using UDP as its transport agent for messaging, the cluster
configuration file, /etc/vx/cvmtab, contains the following fields:
- 3 - Formatted: January 24, 2005
vxclustadm(1M) VxVM 3.5 vxclustadm(1M)
1 Jun 2002
clustername cluster_name
port vxconfigd port_number
port vxkmsgd port_number
node node_ID name name_on_local_net
timeout timeout_value
...
The recommended port numbers for the vxconfigd and vxkmsgd daemons are
4500 and 4501, but any available port numbers greater than 1024 are
also acceptable.
name_on_local_net is the node's IP address or resolvable host name on
the cluster's private network.
timeout_value is the timeout value in seconds. The clustering
functionality of VxVM uses this value during cluster reconfiguration.
The appropriate value to use depends on the number of nodes in the
cluster and on the size of the shared disk group configuration. In
most cases the value of 200 seconds is sufficient but this may need to
be increased for larger configurations.
Comment lines in the file start with a #.
If GAB is being used as the transport agent for messaging, fields
relating to port numbers and local network names are not required:
clustername cluster_name
node node_ID name
timeout timeout_value
...
For a cluster running in the VCS environment, VxVM obtains information
about the cluster from the VCS cluster configuration file
(/etc/VRTSvcs/conf/config/main.cf). Cluster-specific information may
be appended to this file by running the vxcvmconfig command. For more
information refer to the VERITAS Cluster File System Installation and
Configuration Guide.
EXAMPLES [Toc] [Back]
A cluster consisting of four nodes, named node0, node1, node2 and
node3, operates without a cluster monitor, and has the following
cvmtab file when UDP is used as the transport agent for messaging:
# ClusterName
clustername CVM1
# Daemon port numbers
port vxconfigd 4500
port vxkmsgd 4501
# NodeID Nodename Localname
node 0 node0 node0_p
node 1 node1 node1_p
- 4 - Formatted: January 24, 2005
vxclustadm(1M) VxVM 3.5 vxclustadm(1M)
1 Jun 2002
node 2 node2 node2_p
node 3 node3 node3_p
# Timeout value
timeout 200
If GAB is used as the transport agent for messaging, the cvmtab file
only needs to contain the following information:
# ClusterName
clustername CVM1
# NodeID Nodename
node 0 node0
node 1 node1
node 2 node2
node 3 node3
# Timeout value
timeout 200
The following commands bring up the cluster on node1 using messaging
over UDP:
vxclustadm -m nocm startnode
vxclustadm -s 1 membership 1
If node1 is the first node to join the cluster, it becomes the master
node. The following command confirms that node1 is the master node:
vxdctl -c mode
To join node3 to the cluster, the following command is run on node3:
vxclustadm -m nocm startnode
Next, the following command is run on both node1 and node3 to update
information about cluster membership:
vxclustadm -s 2 membership 1 3
To confirm that node3 is a slave node, the following command is run on
node3:
vxdctl -c mode
- 5 - Formatted: January 24, 2005
vxclustadm(1M) VxVM 3.5 vxclustadm(1M)
1 Jun 2002
node1 remains as the master node for its lifetime in the cluster. To
remove node1 from the cluster, the following command is run on node1:
vxclustadm stopnode
After the following command is run on node3, node3 becomes the new
master node.
vxclustadm -s 3 membership 3
For clusters running GAB outside of the VCS environment, the following
command starts the cluster with GAB messaging:
vxclustadm -m gab -t gab startnode
Either of the following commands starts the same cluster with default
UDP messaging:
vxclustadm -m gab startnode
vxclustadm -m gab -t udp startnode
NOTES [Toc] [Back]
vxclustadm does not ensure the consistency of cluster membership
information; you must make sure that the same cluster membership
information is provided to all the nodes.
SEE ALSO [Toc] [Back]
vxconfigd(1M), vxdctl(1M), vxintro(1M)
VERITAS Volume Manager Administration Guide
VERITAS Cluster File System Installation and Configuration Guide
- 6 - Formatted: January 24, 2005 [ Back ] |
