idscheck 2.3.0

A command line tool for IDS student to check GPU allocation

IDS Check

A library for docker users to check the GPU and CPU utilization by simple command line commands.

Install

Use pip3 to install the library idscheck:

pip3 install idscheck

How to Use

After install, you can just use the following easy commands to check the status of GPU and CPU.

Most efficient way

Just use the ids command to check who is using the GPU and the email of them:

ids

Example Outputs are:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.73.08    Driver Version: 510.73.08    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA A100-SXM...  On   | 00000000:07:00.0 Off |                    0 |
| N/A   30C    P0    63W / 400W |   3823MiB / 40960MiB |     27%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   1  NVIDIA A100-SXM...  On   | 00000000:0F:00.0 Off |                    0 |
| N/A   31C    P0    86W / 400W |  13140MiB / 40960MiB |     21%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   2  NVIDIA A100-SXM...  On   | 00000000:47:00.0 Off |                    0 |
| N/A   27C    P0    52W / 400W |      2MiB / 40960MiB |      0%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   3  NVIDIA A100-SXM...  On   | 00000000:4E:00.0 Off |                    0 |
| N/A   32C    P0    84W / 400W |  26357MiB / 40960MiB |    100%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   4  NVIDIA A100-SXM...  On   | 00000000:87:00.0 Off |                    0 |
| N/A   34C    P0    66W / 400W |  30134MiB / 40960MiB |     25%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   5  NVIDIA A100-SXM...  On   | 00000000:90:00.0 Off |                    0 |
| N/A   62C    P0   296W / 400W |  38917MiB / 40960MiB |     97%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   6  NVIDIA A100-SXM...  On   | 00000000:B7:00.0 Off |                    0 |
| N/A   33C    P0    76W / 400W |   9309MiB / 40960MiB |      0%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   7  NVIDIA A100-SXM...  On   | 00000000:BD:00.0 Off |                    0 |
| N/A   63C    P0   270W / 400W |  38865MiB / 40960MiB |     93%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+

+-------+---------+--------------+-----------------+-----------+---------------------------+
| GPUID |   PID   | Process Name | Used GPU Memory |    User   |      Email Address        |
+-------+---------+--------------+-----------------+-----------+---------------------------+
|     0 | 1302626 | python_naibo |         3821MiB |   naibo   | [e-mail] |
|     1 | 1584227 | python_naibo |         3831MiB |   naibo   | [e-mail] |
|     4 | 1570396 | python_naibo |         3777MiB |   naibo   | [e-mail] |
+-------+---------+--------------+-----------------+-----------+---------------------------+

Check GPU Utilization Info

Just like nvidia-smi, use idsgpu or ids gpu to check GPU utilization.

idsgpu

or

ids gpu

Example output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.73.08    Driver Version: 510.73.08    CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA A100-SXM...  On   | 00000000:07:00.0 Off |                    0 |
| N/A   44C    P0   109W / 400W |   4606MiB / 40960MiB |     35%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
|   1  NVIDIA A100-SXM...  On   | 00000000:0F:00.0 Off |                    0 |
| N/A   35C    P0   126W / 400W |   3829MiB / 40960MiB |     22%      Default |
|                               |                      |             Disabled |
|   2  NVIDIA A100-SXM...  On   | 00000000:BD:00.0 Off |                    0 |
| N/A   58C    P0   263W / 400W |   3427MiB / 40960MiB |     91%      Default |
|                               |                      |             Disabled |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A    423426      C   python_naibo                     3425MiB |
|    0   N/A  N/A    432801      C   python_naibo                     1391MiB |
|    1   N/A  N/A    430531      C   python_naibo                     3827MiB |
|    2   N/A  N/A    882885      C   python                          26355MiB |
+-----------------------------------------------------------------------------+

Check Process

Just like the top commands, use idstop or ids top to check all the running processes:

idstop

Or you can filter some information, e.g, check who is running process with PID 423426:

ids top | grep 423426

Example Output:

4189111 naibo     20   0   19.8g   6.0g  43720 S  2270   0.6 120124:48 python

Check User Information to contact them

Use idsquery or ids query to get all the user's email addresses to contact them if you want them to leave some GPU/CPU resources for you:

idsquery

Example Output:

+--------------+---------------------------+
|  Username    |       Email Address       |
+--------------+---------------------------+
|   naibo      | [e-mail] |
|   xiaoming   |     [e-mail]     |
+--------------+---------------------------+

Check All processes' information

Use idstopall or ids topall to get all the processes' information, including the root user of the server.

Notify other users

If you want to use GPU but you cannot find 2 GPUs at your server, you can use the idscheck tool to notify and make the users who occupied more than 2 GPUs to free their GPUs for you, the instructions are as follows:

1. Upgrade/Install the idscheck tool via pip3:

pip3 install idscheck –upgrade

2. Check the GPU utilization first to make sure you cannot find any GPU available:

ids 3. (Optional) To check CPU utilization, you can use:

Idstop

4. Then you can notify the users by “idsgpu” or “idsnotify”:

Idsgpu

5. Then you and the over-used users will receive an email to notify them to free their GPUs within 24 hours, and they will not see your email address so that you don’t need to worry about the possible friction.
6. Then at most 24 hours later, you will receive an email that says you can now use GPU, because when exceed 24 hours and they still don’t kill their processes, the system will automatically kill all their processes.
7. If they really need to use the GPUs, such as they are meeting the paper deadline, they will send me an email and I will tell you this situation, and let you know when is the earliest time you can get access to the GPU, if you agree that’s good, if you don’t agree, we can build an anonymous group to negotiate.