Set up Cluster Foundation (CF) and CIP by using the CF Wizard of Cluster Admin. The designsheets for the CF and CIP setup are "Setup (initial configuration)" of PRIMECLUSTER Designsheets.
For details on the setting, see "1.1.4 Example of creating a cluster" in "PRIMECLUSTER Cluster Foundation (CF) Configuration and Administration Guide."
Setup item | Description | Described in designsheets |
|---|---|---|
Cluster name | Define the name of the cluster systems. | "Cluster name" |
Cluster nodes | Select the nodes that will construct a cluster system. | "Node name (uname-n)" for "Node 1", "Node 2", "Node 3" and "Node 4" |
CF node names | Set the names of the nodes that construct the cluster. | "CF node name" for "Node 1", "Node 2", "Node 3" and "Node 4" |
Cluster interconnects | In each node of the cluster, determine the network interface to be used in CF inter-node communication. A representative network interface is the Ethernet device. | "Path 0 NIC name" and "Path 1 NIC name" for "Node 1", "Node 2", "Node 3" and "Node 4" |
IP interconnects | Optional. This setup allows you to operate CF over IP. This setting is required in a cloud environment. | "IP interconnect setup" |
CIP subnets | Set the following items, and set the IP address used by CF:
Set the above so that the network segments will be different from those for the IP addresses of all the network interfaces on the OS, including the network interfaces used for IP interconnects. | "Number of subnets", "Subnet IP", and "Net mask" for "CIP" |
Usage confirmation of CF remote services | Check whether the following functions are to be enabled:
With the default settings, these services are "Not selected." To use RMS, you need to select at least one function. | "CF remote service usage" |
Cluster Integrity Monitor (CIM) configuration | Set the nodes to be monitored by CIM. | "Node in CF quorum set" for "Node 1", "Node 2", "Node 3" and "Node 4" |
Note
Node names of the cluster nodes are automatically input to "CF node names." The CF node name must be within 11characters.
When constructing multiple clusters, and if any of NICs used in different clusters exist on the same network, consider isolating the network either physically or by VLAN. If you need to make them identical after consideration, specify a different name per each cluster, such as including the node name in the cluster name.
If you enable any one of the CF remote services, do not connect the following systems in the same cluster interconnect:
Systems that have a security problem
Systems in which cluster interconnects are not secured
Hereinafter, the CF remote services (CFCP and CFSH) must be enabled. To enable this function after configuring CF, add the following definition to the /etc/default/cluster.config file and execute cfset -r.
CFCP "cfcp" CFSH "cfsh"
If the CF and CIP configuration fails, check the following:
The cluster interconnect is incorrect.
The network interface that is used for the cluster interconnect is not activated.
After the CF setup is completed, "SF Wizard Startup Check" pop-up window is displayed. Select [No]. SF Wizard is not available in this version. Set up the SF according to the instructions in "5.1.2 Setting up the Shutdown Facility."
To share a NIC with the administrative LAN and the cluster interconnect, see "1.1 CF, CIP, and CIM configuration" in "PRIMECLUSTER Cluster Foundation (CF) Configuration and Administration Guide."
When the bonding device is used for the cluster interconnection, only mode=1(active-backup) can be used.
For the cluster interconnect, it is recommended to use the physically independent and dedicated network. If the network is shared with other communications, a heartbeat failure may be detected due to the temporary network overload. Before the actual operation, test the communication status under the actual network overload and make sure that a heartbeat failure is not detected. If the failure is detected, refer to "11.3.1 Changing Time to Detect CF Heartbeat Timeout" and tune the cluster timeout value.
When configuring the cluster system using the extended partitions in PRIMEQUEST 3000 series (except B model), up to 4 nodes can be supported per cluster system.
You need to use CF over IP in a cloud environment. In addition, you need to configure CIP because such as RMS uses it for inter-node communication in the cluster. CF over IP (IP interconnect) and CIP are different features. For more information, see "1.1.1 Differences between CIP and CF over IP" in "PRIMECLUSTER Cluster Foundation (CF) Configuration and Administration Guide."
When you use VLAN interfaces as interconnects, see "1.1.6 Example of CF configuration by CLI" in "PRIMECLUSTER Cluster Foundation (CF) Configuration and Administration Guide" and configure CF using the CLI.
Note
In the case of the single-node cluster operation
In a physical environment, a KVM environment, or a VMware environment, make sure to set up the cluster interconnect.
For the network interface specified to the cluster interconnect in a physical environment, a KVM environment, or a VMware environment, select a dedicated network interface device listed in the above table.
In a cloud environment, perform the following procedure to create a dummy interface and set the dummy interface as an interconnect. In the following procedure, the interface name of the dummy interface is dummy0.
Execute the following command.
# nmcli connection add type dummy ifname dummy0 ipv4.method manual ipv4.address <IP address>/<prefix length>
Execute the following command to make sure that the dummy interface has been created.
# ip addr show dummy0After the CF setup is completed, "SF Wizard Startup Check" pop-up window is displayed. Select [No] since the setting of the shutdown facility is unnecessary.
Following messages of the shutdown facility and RMS are output, however, this is no problem, since the setting of the shutdown facility is not performed.
Messages of the shutdown facility:
fopen of /etc/opt/SMAW/SMAWsf/rcsd.cfg failed, errno 2 Could not correctly read the rcsd.cfg file
Messages of RMS:
(SCR,26): ERROR The sdtool notification script has failed with status 1 after dynamic modification.
See
For information on the corrective action to be applied when the setting of the cluster interconnect fails, see "Chapter 8 Diagnostics and troubleshooting" in "PRIMECLUSTER Cluster Foundation (CF) Configuration and Administration Guide."