| Shortcut | Action |
|---|---|
| Tab | Autocomplete the path under the cursor |
| Tab Tab (double tap) | Show options for autocomplete |
| ↑ | Recall the previous command (press multiple times as needed) |
| Esc + . | Recall the last argument of the last command (press multiple times as needed) |
| Ctrl + R | Search through your command history (press multiple times as needed) |
| Ctrl + A | Move the cursor to the beginning of the line |
| Ctrl + E | Move the cursor to the end of the line |
| Ctrl + D | Delete the character under the cursor |
| Ctrl + K | Cut the line from the cursor to the end |
| Ctrl + U | Cut the line from the beginning to the cursor |
| Ctrl + Y | Paste what you previously cut |
| Ctrl + C | Kill the current running command |
Use symlinks to save time typing paths to directories or files you use frequently. For example, when you ssh into the HPC, your working directory is initially your home directory (/home/uniqname/). But we keep our project files in a shared lab drive: /nfs/turbo/schloss-lab/. You may be actively working on a project in /nfs/turbo/schloss-lab/uniqname/project_name/ – that can be long and hard to remember. To save time, you can create a symlink to your the project in your home directory like so:
ln -s /nfs/turbo/schloss-lab/uniqname/project_name/ /home/uniqname/project_name/
Then when you log in, you can simply type cd project_name to get to the project directory without typing the whole path. You can use a symlink wherever you would use a normal path, such as in ls, cd, cp, or reading/writing files in scripts.
When you run the same commands repeatedly, you might like to
create aliases to save yourself time typing long commands with lots of options.
Add your aliases to your ~/.bashrc or ~/.zshrc.
For example, typing snakemake -c 1 every time you want to run a snakemake workflow gets tiresome.
Instead, you can define an alias to assign this command to smk.
In your .zshrc file, add this line:
alias smk="snakemake -c 1"
Then, you can run smk to call snakemake with 1 core.
You can also use more cores or specify any other additonal snakemake options with smk:
smk -c 8 --configfile config/test.yml
Here are more aliases that we’ve found to be helpful:
# run snakemake with 1 core.
alias smk="snakemake -c 1"
# print the git commit history.
alias glo="git log --oneline --abbrev-commit --all --graph --decorate --color"
# show the jobs in the queue on GreatLakes or other HPC running SLURM.
# this modifies the default format to make it more readable.
alias sqs='squeue -u $USER -o "%.18i %50j %.2t %.6M %R"'
If you frequently login to servers with ssh, you can save time by creating a config file.
Create a file called ~/.ssh/config with your preferred text editor and add entries for each server like so:
Host host_nickname
Hostname server.address.com
User your_username
Port 8674
Then, instead of typing ssh -p 8674 your_username@server.address.com, you can type ssh host_nickname. This also works for scp and rsync!
An entry for Great Lakes looks like this:
Host gl
Hostname greatlakes.dclab-skku.skku.edu
User uniqname
Then to login, type ssh gl.
Sometimes, computers run very slow because there are processes running that take up too much CPU. This can happen on shared HPC resources (e.g. Flux, Great Lakes) when users run scripts on the login node instead of properly submitting jobs to the job scheduler. You can use a command called top to view what processes are running on the node you’re logged into. Normally, the output of top will look something like this:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
----------------------------------------------------------------------------------------------------------
23047 user1 20 0 107972 680 572 S 12.1 0.0 0:09.37 wc
16222 user2 39 19 7925204 329232 6472 S 1.6 0.7 1166:53 java
23170 me 20 0 162828 3136 1600 R 1.6 0.0 0:00.89 top
1 root 20 0 193772 6108 2284 S 0.0 0.0 88:18.86 systemd
Each row is a process. You can see the uniqname of the person running the process in the USER column and how much compute resources each process is using in the %CPU column. In this case, all of the processes have a low %CPU, so everything is fine here.
When someone runs scripts that they shouldn’t be running on the login node, it may look something like this:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
----------------------------------------------------------------------------------------------------------
23050 user1 20 0 107975 684 575 S 175.3 0.5 3:33.54 python
23049 user1 20 0 107974 683 574 R 174.4 0.6 3:30.18 python
23048 user1 20 0 107973 682 573 S 173.5 0.5 3:28.22 python
23047 user1 20 0 107972 681 572 S 173.2 0.5 3:32.50 python
16222 user2 39 19 7925204 329232 6472 S 1.6 0.7 1166:53 java
23170 me 20 0 162828 3136 1600 R 1.6 0.0 0:00.89 top
1 root 20 0 193772 6108 2284 S 0.0 0.0 88:18.86 systemd
Here, user1 has multiple processes running that are taking up over 100% CPU. As a result, there are few resources left for other users who need to use the login node to do simple tasks like moving files, editing scripts, or checking on jobs. Since the USER column corresponds to the user’s uniqname, you can send them a polite email asking them to kill their processes on the login node and submit them to the job scheduler instead.
Sorted from nearest to farthest from 1526 MSRB I.