跳转至

Hpc guide overview

Login system

With vpn enabled, start a command line tool such as terminal (mac) or cmd (windows): Enter the following command to log in to the server: 

1
ssh p-hwc-sched-1.biolab.ekicloud.com
When you log in for the first time, you will receive the following prompt: 
1
2
3
4
The authenticity of host 'sched (10.27.12.5)' can't be established.
ECDSA key fingerprint is SHA256:ZNXGHp3Gi3lziOdaCUDRinH4qSt7KyHZyqtRp0C49a8.
ECDSA key fingerprint is MD5:49:91:8c:9d:0b:ac:44:fb:75:4c:2b:f3:43:4f:51:4f.
Are you sure you want to continue connecting (yes/no)?
After typing yes you will be prompted for a password: 
1
2
Warning: Permanently added 'p-hwc-sched-1.biolab.ekicloud.com,10.27.12.5' (ECDSA) to the list of known hosts.
zhenghy@p-hwc-sched-1.biolab.ekicloud.com's password:
Enter your password to complete the login, and the welcome interface will pop up after login: 
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
Last login: Mon Apr 26 10:24:21 2021 from 10.3.1.102
(C) 2021 Beijing Ekitech Co. Ltd.

===============================================
 - Hostname............: p-hwc-sched-1.biolab.ekicloud.com
 - Disk Space..........: 64G remaining
===============================================
 - CPU usage...........: 0.00, 0.01, 0.05 (1, 5, 15 min)
 - Memory used.........: 832 MB / 7820 MB
 - Swap in use.........: 0 MB
===============================================

Warning

The login node itself cannot perform tasks, please do not perform tasks directly on the login node

File storage and work path

Generally, you can store your software and execute your commands in two places, your home directory "/mnt/home/\(user" and your working directory "/mnt/scratch/\)user" The data files you store on obs are stored in the read-only directory "/mnt/bucket".

/mnt/home/$user

This is your personal folder where you can keep things like frequently used software and reference data. You can enter this directory with the following commands: 

1
cd

1
cd ~

1
cd /mnt/home/$user

Warning

The files stored here will incur a fee depending on the total size, so it is not recommended to store a large number of files here for a long time.

/mnt/scratch/$user

This is a shared working directory. You can perform tasks or store temporary files here. The files here will be cleaned up regularly. Therefore, please store the data that needs to be stored for a long time in obs in time.

You can enter this directory with the following command 

1
cd /mnt/scratch/$user

Warning

The cleaning cycle of the files stored is 30 days. After the files are cleaned up, they cannot be recovered. Please transfer important files in time.

/mnt/bucket

This is the address of your organization's obs network disk. You can view and read files stored in obs in this read-only directory.

Submit tasks

There are two main ways to submit tasks to the computing node at the login node, ekbatch and ekrun.

ekbatch will help you generate a task script and automatically submit the task for you. The task script and running results will be stored in "/mnt/scratch/$user/ekbatch" Enter the ekbatch command There will be the following prompts： 

1
2
3
Slurm Job Script Generator for hpc-class
Please set memory for this job.
  Notes: maximum memory you are allowed is 32 >
Enter the number of memory (such as 1G) as needed. If you do not enter the unit, the default unit of MB will be used, and then you will be prompted to enter the required number of CPUs 
1
2
Number of processor cores per node
  Notes: maximum processor cores you are allowed is 16 >
Enter the number of cpus as needed (like 1) Next, enter the task you want to perform 
1
2
Specify the executable of your application.
 >
Finally, enter the name of this task in order to view the results 
1
Specify a name for your job.
The complete process is as follows 
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
❯ ekbatch
Slurm Job Script Generator for hpc-class
Please set memory for this job.
  Notes: maximum memory you are allowed is 32 >1G
Number of processor cores per node
  Notes: maximum processor cores you are allowed is 16 > 1
Specify the executable of your application.
 > hostname | wait 10 | hostname
Specify a name for your job.
 > show
your job begins running
Submitted batch job 311482

The ekrun command is used to perform tasks that require real-time feedback. You only need to add ekrun to your command to submit a task to the supercomputing system with one cpu and two gigabytes of memory and get feedback in real time. Like ekrun your job;

You can also enter interact anywhere in the command to manually enter information such as the number of memory required and the number of cores to compute. Such as ekrun interact hostname. Executing the above command will get the following result: 

1
2
3
4
5
6
❯ ekrun interact hostname
Please set memory for this job:
 > 1
Number of processor cores per node
  Notes: maximum processor cores you are allowed is 32
 >  1

Inquire your tasks

You can query your current task progress by entering ekqueue. The query result is generally as shown below: 

1
2
JOBID PARTITION                                               NAME     USER ST       TIME  NODES NODELIST(REASON)
311482     guest                               show_20210426-142156  xhliang CF       1:29      1 node-g-202

回到顶部