Utilities

This section describes:

Cluster executable

You may use the cluster.exe utility program to start multiple processing nodes on the local machine. In this section there is information on:

Cluster Dialog Box

Go to File > Tools > Cluster Node Managementto open the Cluster Node Management dialog box.

Select the Work Directory and Process ID of the desired Cluster processes. Click Browse to pick a folder from the local system, or remote network.
Under Process List, enter the applicable Process IDs.
In Time to Wait for Response, enter the number of seconds CUBE Cluster should wait for a response from a node before determining that the node is unavailable.
Click List Nodes to return the status of all processing nodes within the chosen Work Directory.
Click Start Nodes to start all listed processing nodes (remote or local). Or, click Start Selected to only start the selected node(s).
Click Close Nodes to close all listed processing node (remote or local). Or, click Close Selected to only close the selected node(s).
Check Hide New Nodes to minimize cluster nodes to the system tray. The nodes will be hidden after they are Started. This is the same as clicking Hide on the associated Voyager instance. Hiding is useful when the user wishes to minimize window clutter.

Running Cluster from the Command Line

You can also run this program from a Windows command line. Therefore, you can run the program from either a .bat batch file or from a CUBE Voyager * command call. The syntax for the command line is:

Cluster [ProcID] [ProcList] [Start/Starthide/Close/List]
[Exit]

These command line options correspond to the Cluster dialog box shown above.

Starting processing nodes in CUBE Cluster

Prior to starting a distributed-processing run using CUBE Cluster, individual instances of CUBE Voyager must be running and set to wait mode with the appropriate script file name and work directory for the run. A CUBE Cluster processing node corresponds to a single, properly configured instance of CUBE Voyager running and waiting. Processing nodes can either be additional processors on the local computer (local nodes) or processors on remote computers connected to the main computer over a local network (remote nodes). You can start an instance of CUBE Voyager manually by running the Voyager.exe program in the CUBE Voyager program directory and setting the script and working directory, or by running the Cluster.exe program as described in Cluster executable.

You cannot start instances of CUBE Voyager on remote nodes over a network. You must start remote nodes directly on the remote machines. Typically, you must physically go to each machine and run either the Voyager.exe or the Cluster.exe program on that machine. Alternatively, you can use the COMMPATH keyword to send information to active instances of CUBE Voyager, which are open and running on remote processing nodes. Use COMMPATH to send changes in script and working directory names to waiting instances of CUBE Voyager without having to physically visit each machine and make changes manually.

You use COMMPATH only for initial communication with the node. The node will switch its work directory to be the same as the main process before running the multistep distributed process. After completing the steps, the node reverts to waiting for the communication file in the COMMPATH directory. Bentley recommends starting an instance of CUBE Voyager on each remote processor that you will use, setting the work directory to a common path like z:wait, and leaving them run. Then, you can use this value as the COMMPATH keyword value.

Utility functions

A number of utility functions were added to the CUBE Voyager system to allow more flexibility in performing distributed processing:

FilesExist(‘file1|file2|file3…’) — This function will check for the existence of one or more files. The function takes one string argument (constant or variable) and if there are more than one file to check, they are put into one string and separated by ‘|’. The return value is the number of files that exist. If none of the files exist, then the return value will be zero. This can be use instead of the Wait4Files command if you only want to check if a node is done but don’t want to wait for it to get done.
NumReadyNodes(‘ProcessID’,’ProcessList’) — This function will check for availability of a list of CUBE Cluster nodes. The second argument is a string with the list of process numbers to check. The return value is the number of ready nodes. For example, NumReadyNodes(‘TestDist’,’1-5,10,15-20’) or NumReadyNodes(MyProcess,MyProcessList)
FirstReadyNode(‘ProcessID’,’ProcessList’) — This function will check for availability of a list of CUBE Cluster nodes and return the process number of the first available node. The second argument is a string with the list of process numbers to check. The processes are checked in the input order and can go from low to high or high to low so if the list is ‘6-2’ and all processes (2-6) are available, the result will be 6. For example, FirstReadyNode(‘TestDist’,’1-5,10,15-20’)

The following is an example for using the NumReadyNodes function when there are two separate groups of CUBE Cluster nodes that may be available to participate in a DP run, the script wants to select the one with the most available nodes:
```
IF (NumReadyNodes(‘DP1’,’1-10’) > NumReadyNodes(‘DP2’,’11-20’))
MyProcessID=’DP1’
MyProcessList=’1-10’
ELSE
MyProcessID=’DP2’
MyProcessList=’11-20’
ENDIF
RUN PGM=Matrix
DistributeIntraStep ProcessID=@MyProcessID@, ProcessList=@MyProcessList@
```