Login
Working with computer clusters takes place remotely via the Internet by connecting to a dedicated access server (login node), where the Centos7 Linux operating system with special cluster and task management tools is installed. After connection, you’ll be able to use the command line with Unix commands. You can also call graphical tools/windows from the command line. For using a cluster in graphical mode (X Windows or Remote Desktop), see section Cluster user interfaces.
Command line access parameters
Primary server address: ui-1.hpc.rtu.lv
Spared (ui-1 copy): ui-2.hpc.rtu.lv
Protocol: SSH
Port:22
The user name and password you received when you registered for the cluster access are used for authentication.
Access is provided from any IP address.
Access tools
The connection to the cluster is via a Secure Shell (SSH) connection. For Windows operating system, you can use PuTTY software (download it from: http://www.putty.org/). The parameters you should enter are given above in the Command line access parameters section. The example is shown in the figure below.
After you open the connection, a user name and password are required. After entering them, it is possible to use a remote terminal with Unix commands.
When you use a Linux or MacOS operating system, you can connect to a cluster with SSH by opening the command line (terminal) and running the following command:
ssh -p [port] [username]@ui-1.hpc.rtu.lvFor example:
ssh -p 22 lauris@ui-1.hpc.rtu.lv Command execution on an access server (including opening the application GUI) does not yet ensure that the task will automatically run on computing nodes. This requires you to use the task management tools described in section Task Management.
We invite to use the access server for copying files to/from cluster, preparing tasks, putting tasks in a queue, monitoring, testing short (<5 min.) tasks, and compiling on a few cores.
User workspace
Each user has its own workspace, where they can store files related to the jobs. By accessing the command line of the login node, the user is automatically directed to it’s work directory.
/home/usernameDirectory is common to all cluster nodes and access server (NFS – Network File System), therefore no copying of files to or from the execution node or between the nodes is necessary. It means if you copy a file on the login node, it will be available on other nodes and all changes made to files on the login node will be also seen on other nodes.
In their working area, the user can also install/compile the application software if the installation process does not require administrator (root) rights.
Copying of files
In order to copy files from your computer to a cluster access server easily, MS Windows users can use, for example, WinScp or Far file manager. WinScp can be downloaded here. Connection is similar to that of PuTTY.
On the left, you’ll see your computer files, and the right – work area in the cluster. You can drag files with a computer mouse from one window to another:
On a MacOS or Linux operating system, use a command-line SCP command or any convenient graphical tool. An example of copying a file from a Linux command line:
scp –r my.file username@ui-1.hpc.rtu.lv:Using Application Modules
Users can use modules (Environment Modules) to prepare the cluster environment for different application program, compilers, and using libraries. This option makes it easy for users to use different versions of the same program, use centrally installed tools, rather than installing/compiling everything on your own. The module is loaded with a command:
module load [module_name]The team sets the path (PATH) to executable files, libraries, and other required environment variables for the respective software. To get a list of all possible modules (application software and tools), follow the steps:
module avail
Loading modules is usually not required for tools included in Linux distributive
The module must be loaded on the node where the program starts. If the task is executed on the computing node/s, the module must be loaded into it well.
Find out the downloaded modules:
module listTo reset the module:
module unload [module_name]More about using modules: http://modules.sourceforge.net/
Personal modules
If a user compiles software in his/her work area, it is also possible to create their own modules. Briefly:
- Create a directory
/home/[lietotāja vārds]/privatemodules- Insert your module into this directory. Samples:
/opt/exp_soft/modulefiles- Always activate the private module directory when starting (you can type in this:
~/.bashrcwhen you run a command
module load use.own- Then load your own module
module load [module_name]