The Settings Menu

This widget is where you select the internal cluster network interface used by the Beowulf server daemon. It defines the communication path between the master node and all the compute nodes in the cluster. This option is stored in the config file as "interface".

This widget displays the IP address of the master node on the internal cluster network interface. This is the address used by the compute nodes when communicating back to the master node. This address must be in the same subnet, but may not fall in the IP address range used for the compute nodes (see below). This setting, along with the net mask, is typically configured by the underlying Linux system based on the contents of the /etc/sysconfig/network-scripts/ifcfg-eth0 file.

This widget displays the network address mask configuration of the internal cluster network interface. This setting is read from the current interface state.

This widget is where you specify the range of IP addresses assigned to the compute nodes on the internal cluster network interface. The maximum number of compute nodes in the cluster is defined by this inclusive range (the range includes the upper address). This option is stored in the config file as "iprange" and the computed value "nodes". BeoSetup assists in setting a valid range by using the net mask value to limit which elements of the addresses may be modified.

These radio buttons let you change the node assignment policy used for assigning node numbers and IP addresses to new nodes as they join the cluster. There are four options:

• Off — The operator can manually configure all new node addresses.

• Insert — New nodes will be assigned to the first available node number. Nodes are numbered sequentially starting from 0.

• Append — New node IP addresses will be appended to the end of the Configured Nodes list box. This option is effectively the same as Insert if the final node number has been assigned.

• Locked — This locks in the current node order and IP addresses. Both Beoserv and BeoSetup will be prevented from adding or changing node IP addresses or node order.

When this box is checked, node information will also be stored in the specified directory, with one file written per node. The directory /etc/beowulf/nodes/ is the default.

These radio buttons let you specify the type of file system check performed during boot. This option is stored in the config file as fsck; there are three possible values:

• Never — No file system check will be performed.

• Safe — The file system check gives up if any "bad" errors are encountered.

• Full — The system will try to answer "yes" to all the "Should I fix?" questions that may occur during the file system check.

These radio buttons let you specify the type of file system create performed during boot. This option is stored in the config file as mkfs; there are three possible values:

• Never — No file system create will be performed.

• If check fails — A blank file system is recreated if the file system check fails.

• Always — The file system is recreated during every boot. If you select this option, the Check (e2fsck) options will be disabled, and the file system check will be skipped.

This checkbox should be selected when the compute nodes must boot from a different network than they use for run-time operation. Input the network interface name, or choose the down arrow for other network selections. The BeoPXE Server will use the specified network interface for PXE booting rather than the cluster network interface. This alternate network interface should be connected to the booting interface for the compute nodes.

Following are three examples of configurations that require an alternate PXE interface:

• When compute nodes have a Fast Ethernet network interface that supports PXE booting and a Gigabit Ethernet interface without PXE support. The Gigabit Ethernet is the desired operational network.

• When compute nodes have an Ethernet network interface that supports PXE booting and a specialized cluster NIC (Myrinet, Infiniband, etc.).

• When compute nodes are RLX compute blades and the only network interface that supports PXE booting is connected to an internal repeater. Using an Ethernet repeater is very slow during normal operation, so the cluster interfaces should be connected to an Ethernet switch, while the master should use a separate network interface connected to the internal RLX repeater. The second interface and the internal repeater are used only for booting.

When this box is checked, BeoSetup limits network use of booting machines to avoid timeouts. You can input a value or use the increment/decrement arrows to select the desired PXE service bandwidth limit in Mbps.

There are 8 columns that can be displayed by checking the corresponding box:

• ID — The Node ID number (required) will be displayed.

• Eth HW Address — The node's unique hardware address (MAC address) will be displayed.

• IP Address — The IP address assigned to the node will be displayed.

• State — the current node status ( such as Down, Up, Rebooting, etc.) will be displayed.

• User — The username of the user who controls the node will be displayed.

• Group — The groupname of the user group that controls the node will be displayed.

• Mode — The execute permissions (user, group, or all) will be displayed.

• Notes — The administrative notes will be displayed.

You can select a highlight color for each of 9 node states:

• Down — Display color for nodes that are down.

• Unavailable — Display color for nodes that are administratively unavailable.

• Halt — Display color for nodes that have been commanded to halt.

• Boot — Display color for nodes that are booting.

• Up — Display color for nodes that are fully operational.

• PowerOff — Display color for nodes that have been commanded to power off.

• Error — Display color for nodes that encountered an error while booting.

• Reboot — Display color for nodes that are rebooting.

• No IP assigned — Foreground color for nodes without an IP address.

If you want to change a color assigned to any of the node states, click the color box next to that state; the custom color wheel panel will open. You can select a custom color by moving the circle within the color wheel, by moving the sliders, or by entering values in the fields. This color wheel is based on the RGB (red-green-blue) color model; you can also specify the hue, saturation, and value of the color.

Check any combination of the boxes to indicate which of the notes should be visible to cluster users. Three default values (Rack, Chassis, or Blade) are available through the dropdown menus; you can also enter a note directly in the fields.