Getting to a remote computer

ssh is an executable program that runs on your local computer and allows you to connect securely to a remote computer.

ssh programs exist for all major operating systems - Windows, Mac, and linux. Mac and linux come with these commands built-in; Windows needs some help. If you're using a Windows box and are part of UT Austin, Bevoware provides two free ssh programs, "ssh secure client" and "putty". We won't describe how to use these here - you're on your own to get this started.

Let's try it now by entering this:

SSH to access Lonestar at TACC
ssh <your user ID>

When you log in to a linux computer, the operating system checks your login credentials and if they're OK it sets up some configuration for you and then runs a program called a "shell" which acts like your fast-food drive-thru window to the rest of the operating system. You type commands and hit "enter" to send something into the drive-thru window, and then the OS passes output back through the drive-thru window.

Every time you exchange stuff through this window, it's within a context, like one specific drive-thru window at one restaurant. The directory within the file system is one part of that context; the programs and environment variables available to you are other parts of that context. When you log in, the system and shell agree that you'll start off in your home directory on the system.

Exercise: Start another shell to Lonestar

Create a new window that will allow you to have a second login to Lonestar, and again use:

SSH to access Lonestar at TACC again, in a new window
ssh <your user ID>

to connect to Lonestar.

Once you're logged in, issue this command (we'll explain later):

Start this process in a new shell on Lonestar
idev -m 200 -q serial

If you have any errors, try to copy and paste the command from the wiki into the shell.

Please switch back to your first login shell once you've entered the idev command in the second shell.

Essential command-line tricks to look like an expert quickly, or figure out what's going on.

Type as little and as accurately as possible by cheating:

  • Cheat 1: Use "up arrow" to retrieve any of the last 500 commands you've typed. You can then edit them and hit enter (even in the middle of the command) and the shell will use that command.

This taps into a feature of the shell - your history. The command history will print to the screen the last 500 commands you've typed. You can modify this number if you'd like. VERY USEFUL TIP - every so often do history >> what which will write your history to a file called "what". I leave these lying around in directories so I can remember what I was doing, how I generated output data, etc. These can often become the basis for a shell script (we'll get to those). Advanced topic: use history to be super-fast at the command line.

  • Cheat 2: Hit the tab key twice - it's almost always magic. This instructs the shell to try to guess what you're doing and finish the typing for you. On most modern linux shells, it works for commands (like "ls" or "scp") and for completing file or directory names.

This is really useful if you can't remember whether is fastaToFastq or or Fasta2fastq or or something else. It's also helpful for reconstructing directory paths or filenames on-the-fly.

You might find write out a long command with a ton of options in the terminal and then find out that you misspelled something at the very beginning of the line. It can be really annoying to hold down the arrow key to get back to that point.

  • Cheat 3: You can use control-a (holding down the "control" key and "a") to jump the cursor right to the beginning of the line. The omega to that alpha is control-e, which jumps the cursor to the end of the line. Arrow keys work, and control-arrowl will skip by word forward and backward.

Unfortunately, you are pretty much out of luck if you want to jump to the middle of the line. In this case you might want to copy the whole command into a nice text editor on your desktop, change it, and copy it back.

Advanced topic: command line editors.


Type "modu" then hit tab twice - it presents two choices, module and modutil. Type the next character l, hit tab twice and it will complete the rest of the typing. If you hit tab twice again, the OS will show you all the files in your current working directory which doesn't make any sense for the command "module" - it's smart, but not smart enough to figure out that the next word in the command needs to be one of module's built-in commands.

Inline help

Man pages - linux has had built-in help files since the mid-1500's, way before Macs or PCs thought of such things. In linux they're called man pages - short for "manual"; it's not a gender thing (I assume). man intro will give you an introduction to all user commands.


Try "man grep", or "man du", or "man sort" - you'll want these sometime.

Tip: Type the letter q to quit man, j and k/<CR> to move up and down by line, b or spacebar up/down by page. Want to search? Just hit the slash key /, enter the search word and hit enter. These are actually the tools of the less command which man is using.

Basic linux commands you need to know like breathing air

  • ls - list the contents of the current directory
  • pwd - print the present working directory - which restaurant am I at right now - the format is something like /home/myID - just like on most computer systems, this represents leaves on the tree of the file system structure, also called a "path".
  • cd <whereto> - change the present working directory to <whereto> - pick up my drive-thru window (shell) and move it so that I'm now looking thru to the directory <whereto>
    • Some special <wheretos>: .. (period, period) means "up one level". ~ (tilde) means "my home directory". ~myfriend (tilde "myfriend) means "myfriend's home directory".
  • df shows you the top level of the directory structure of the system you're working on, along with how much disk space is available
  • head <file> and tail <file> shows you the top or bottom 10 lines of a file <file>
  • more <file> and less <file> both display the contents of <file> in nice ways. Read the bit above about man to figure out how to navigate and search when using less
  • file <file> tells you what kind of file <file> is.
  • cat <file> outputs all the contents of <file> - CAUTION - only use on small files.
  • rm <file> deletes a file. This is permanent - not a "trash can" deletion.
  • cp <source> <destination> copies the file source to the location and/or file name destination}. Using . (period) means "here, with the same name". * cp -r <dirname> <destination> will recursively copy the directory dirname and all its contents to the directory destination.
  • scp <user>@<host>:<source> <destination> works just like cp but copies source from the user user's directory on remote machine host to the local file destination
  • mkdir <dirname> and rmdir <dirname> make and remove the directory "dirname". This only removes empty directories - "rm -r <dirname>" will remove everything.
  • wget <url> fetches a file with a valid URL. It's not that common but we'll use wget to pull data from one of TACC's web-based storage devices.
Wildcards and special file names.

The shell has shorthand to refer to groups of files by allowing wildcards in file names. * (asterisk) is the most common; it is a wildcard meaning "any length of any characters". Other useful ones are [] to allow for any character in the set <characters>> and {{[] for a range of characters.

For example: ls *.bam lists all files in the current directory that end in .bam; ls [A-Za-z]*.bam does the same, but only if the first character of the file is a capital letter.

Three special file names:

  1. . (single period) means "this directory".
  2. .. (two periods) means "directory above current." So ls -l .. means "list contents of the parent directory."
  3. ~ (tilde) means "my home directory".

Scavenger hunt practice; on Lonestar issue the following commands:

Play a scavenger hunt for more practice
cp -r /corral-repl/utexas/BioITeam/linuxpractice .
cd linuxpractice
cd what
cat readme

and follow the instructions. Hints: use <tab><tab> to fill in filenames as much as you can.

Use variables to store where you are, move away, and then back. Try this and see if you can figure out what the shell is doing for you:

Practice some linux basics
cd /scratch/01057
cd $here

Learn about these few advanced tricks (by trying, man pages, Google...)

Advanced tricks
pushd / popd
cd -
which <command>

If you've done all those too, you might consider looking over some advanced command-line tool usage

Options: the lifeblood of linux commands

Sitting at the computer, you should have some idea what you need to do. There's probably a command to do it. If you have some idea what it starts with, you can type a few characters and hit tab twice to get some help. If you have no idea, you google it or ask someone else. But soon you want those commands to do a bit more - like seeing the sizes of files in addition to their names.

Most commands in linux use a common syntax to ask more of a command; they usually add a dash "-" followed by a code letter that means "do the basic command, but with a bit more..."

Useful options for ls
ls -l
ls -lh
ls -t

These little toggle-like things are often called "command line switches"; there can be other options, like filenames, that aren't switches.

Almost all commands, and especially NGS tools, use options heavily.

Like dialects in a language, there are at least three basic schemes commands/programs accept options in:

  1. One letter options which can sometimes be combined, or other single options like:
    Examples of different option types
    head -10
    ls -lhtS (equivalent to ls -l -h -t -S)
  2. Word options, like -d64 and -Xms512m in this command, that are never combined (this is the GATK command to call SNPs):
    Examples of word options
    java -d64 -Xms512m -Xmx4g -jar /work/01866/phr254/gshare/Tools_And_Programs/bin/GenomeAnalysisTK.jar -glm BOTH -R $reference -T UnifiedGenotyper -I
    $outprefix.realigned.recal.bam --dbsnp $dbsnp -o $outprefix.snps.vcf -metrics snps.metrics -stand_call_conf 50.0 -stand_emit_conf 10.0 -dcov 1000
    -A DepthOfCoverage -A AlleleBalance
  3. "Long option" forms, using the convention that a single dash - precedes single-letter options, and double dashes- - precede word options, like this command to run the mira assembler:
    Example of long options
    mira --project=ct --job=denovo,genome,accurate,454 -SK:not=8

man pages should detail all options available for a command. Unless there's no man page.

More help please

Sometimes man lets you down - no man page. Don't fret, try one of these:

  1. Just type in the command and hit return - it will usually try to help you.
  2. Type the command followed by one of: -h -help --help -? and may give you some help.
    Sometimes the command by itself will give you short help, and will list the magic option for full help.

First do:

module load blast

Now figure out how to run some kind of blast program on lonestar with options. Hints: try <tab><tab>, man, running some blast command, use options to figure out other options.

I've put nr, nt, and refseq_rna blast databases on Lonestar here:
along with a test sequence: the human JAG1 gene, here:

blastn -query jag1.fa -db /corral-repl/utexas/BioITeam/blastdb/hs37d5.fa -evalue 1e-100 -outfmt 6
But of course you wouldn't run this on the head node - you'd instead enter it into a file called "commands" using a text editor, or do exactly this:

echo "blastn -query $BI/sphsmith/jag1.fa -db \
   $BI/blastdb/hs37d5.fa -evalue 1e-100 -outfmt 6" > commands
/corral-repl/utexas/BioITeam/bin/ -l blast.sge -n blast_jag1 -t 00:30:00 -j commands
qsub blast.sge

Now let's go on to establishing a useful login profile on Lonestar.

  • No labels