File systems and transferring files
- Anna Battenhouse
- Daryl Barth
- Rachael Michelle Cox
File systems at TACC
The first thing you'll want to do is transfer your sequencing data to TACC so you can process it there. Here is an overview of the different storage areas at TACC, their characteristics, and Linux commands generally used to perform the data transfers:
- wget – retrieves the contents of an Internet URL
- cp – copies directories or files located on any local file system
- scp – copies directories or files to/from a remote system
- rsync – copies directories or files on either local or remote systems
Read more about Copying files and directories
| TACC storage areas and Linux commands to access data (all commands to be executed at TACC except laptop-to-TACC copies, which must be executed on your laptop) |
|
Local file systems
There are 3 local file systems available on any TACC compute cluster (stampede3, lonestar6, etc.), and your account has a directory in each of the three.
The 3 file local systems have different characteristics, but all are fast and set up for parallel I/O.
On lonestar6 these local file systems have the following characteristics:
| Home | Work | Scratch | |
|---|---|---|---|
| quota | 10 GB | 1024 GB = 1 TB | none |
| policy | backed up | not backed up, not purged | not backed up, purged if not accessed recently (~10 days) |
| access command | cd | cdw | cds |
| environment variable | $HOME | $WORK (different sub-directory for each cluster) $STOCKYARD (root of the shared Work file system) | $SCRATCH |
| root file system | /home | /work | /scratch |
| use for | Small files such as scripts that you don't want to lose. | Medium-sized files you don't want to copy over all the time. For example, custom programs you install (these can get large), or annotation file used for analysis. | Large files accessed from batch jobs. Your starting files will be copied here from somewhere else, and your final results files will be copied elsewhere (e.g. stockyard, corral, your BRCF POD, or your organization's storage area. |
When you first login, the system gives you information about disk quotas and your compute allocation balance in "SU" (system units).
--------------------- Project balances for user abattenh ---------------------- | Name Avail SUs Expires | Name Avail SUs Expires | | OTH21095 27688 2025-01-31 | DNAdenovo 3000 2024-09-30 | | OTH21164 1010 2025-03-31 | OTH21180 996 2025-03-31 | ------------------------ Disk quotas for user abattenh ------------------------ | Disk Usage (GB) Limit %Used File Usage Limit %Used | | /scratch 0.0 0.0 0.00 0 0 0.00 | | /home1 0.0 11.7 0.02 316 0 0.00 | | /work 169.0 1024.0 16.50 79361 3000000 2.65 | -------------------------------------------------------------------------------
changing TACC file systems
When you first login, you start in your Home directory. Use the cd, cdw and cds commands to change to your other file systems. Notice how your command prompt helpfully changes to show your location.
cdw # cd $WORK cds # cd $SCRATCH cd # cd $HOME
The cd (change directory) command with no arguments takes you to your Home directory on any Linux/Unix system.
The cdw and cds commands are specific to the TACC environment.
Stockyard (shared Work)
TACC compute clusters now share a common Work file system called stockyard. So files in your Work area do not have to be copied, for example from to stampede3 to ls6 – they can be accessed directly from either cluster.
Note that there are two environment variables pertaining to the shared Work area:
- $STOCKYARD - This refers to the root of your shared Work area
- e.g. /work/01063/abattenh
- e.g. /work/01063/abattenh
- $WORK - Refers to a sub-directory of the shared Work area that is different for different clusters, e.g.:
- /work/01063/abattenh/ls6 on lonestar6
- /work/01063/abattenh/stampede3 on stampede3
A mechanism is being developed for purchasing larger stockyard allocations (above the 1 TB basic quota) from TACC.
The UT Austin BioInformatics Team, a loose group of bioinformatics researchers, maintains a common directory area on stockyard.
ls /work/projects/BioITeam
Files we will use in this course are in a sub-directory there. The $CORENGS environment variable set in your login profile refers to this path.
echo $CORENGS ls /work/projects/BioITeam/projects/courses/Core_NGS_Tools
Corral
corral is a gigantic (multiple PB) storage system (spinning disk) where researchers can store data. UT researchers may request up to 5 TB of corral storage through the normal TACC allocation request process. Additional space on corral can be rented for ~$43/TB/year. See https://docs.tacc.utexas.edu/hpc/corral/
A couple of things to keep in mind regarding corral:
- corral is a great place to store data in between analyses.
- Store your permanent, original sequence data on corral
- Copy the data you want to work with from corral to $SCRATCH
- Run your analyses (batch jobs)
- Copy your results back to corral
- Occasionally corral can become unavailable. This can cause any command to hang that tries to access corral data!
Ranch
ranch is a gigantic (multiple PB) tape archive system where researchers can archive data. All TACC users have an automatic 2 TB ranch allocation. UT researchers may request larger (multi-TB) ranch storage allocations through the normal TACC allocation request process.
There is currently no charge for ranch storage. However, since the data is stored on tape it is not immediately available – robots find and mount appropriate tapes when the data is requested, and it can take minutes to hours for the data to appear on disk. The metadata about your data – the directory structures and file names – is always accessible, but the actual data in the files is not on disk unless "staged". See the ranch user guide for more information: https://docs.tacc.utexas.edu/hpc/ranch/.
Once that data is staged to the ranch disk it can be copied to other places. However, the ranch file system is not mounted as a local file system from the stampede3 or ls6 clusters. So remote copy commands are always needed to copy data to and from ranch (e.g. scp, rsync).
About file systems
File systems are storage areas where files and directories are arranged in a hierarchy. Any computer can have one or more file systems mounted (accessible as local storage). The df command can be used on any Unix system to show all the "top-level" mounted file systems. TACC has a lot of temporary file systems, so lets just look at the first 15 and tell df to use "human readable" size formatting with the -h option:
df -h | head -15
The rightmost Mounted on column give the top-level access path. Find /home1, /work, and /scratch and note their Size numbers!
Filesystem Size Used Avail Use% Mounted on devtmpfs 126G 0 126G 0% /dev tmpfs 126G 43M 126G 1% /dev/shm tmpfs 126G 4.1G 122G 4% /run tmpfs 126G 0 126G 0% /sys/fs/cgroup /dev/md127 150G 92G 59G 61% / /dev/sda2 1014M 207M 808M 21% /boot /dev/md126 284G 21G 264G 8% /tmp /dev/md125 8.0G 4.3G 3.8G 54% /var 129.114.40.1:/admin 3.5T 714G 2.6T 22% /admin 129.114.40.7:/home1 7.0T 6.5T 504G 93% /home1 172.29.200.10@o2ib1172:172.29.200.11@o2ib1172:/work 6.8P 2.5P 4.3P 37% /work 129.114.52.169:/corral/main 38P 21P 18P 55% /corral tmpfs 26G 0 26G 0% /run/user/891443 tmpfs 26G 0 26G 0% /run/user/881379
What do we mean by "hierarchy"? The file system hierarchy is like a tree, with the root file system (denoted by the leading / ) as the trunk, sub-directories as branches, sub-sub-directories as branches from branches (and so forth), with files as leaves off any branch.
your TACC Home directory |
|
But everyone has a Home directory, so you must only be seeing a part of the Home directory hierarchy. To see the absolute path of a directory you're in, use the pwd -P command. Note that absolute paths always start with a forward slash ( / ) denoting the root file system.
pwd -P # will show something like this # /home1/01063/abattenh
That shows you that your Home directory (~) is actually 3 levels down in the /home1 hierarchy:
| part of the TACC Home file system using absolute paths |
|
Here's a depiction of the three file systems as seen from your Home directory ( ~ ), showing where the path-valued environment variables represent, and where the three symbolic links (~/CoreNGS, ~/scratch, ~/work) you created in your Home directory point. Notice that both the Work and Scratch file systems have a top-level hierarchy like we saw in Home above.
On many Linux systems, you can use the tree command to view the full file system hierarchy starting from a specified directory:
Staging your data
So, your sequencing center has some data for you. They may send you a list of web or FTP links to use to download the data.
The first task is to get this sequencing data to a permanent storage area. This should NOT be your laptop! corral (or stockyard) is a great place for it, a BRCF pod, or a server maintained by your lab or company.
Here's an example of a "best practice". Wherever your permanent storage area is, it should have a rational sub-directory structure that reflects its contents. It's easy to process a few NGS datasets, but when they start multiplying like tribbles, good organization and naming conventions will be the only thing standing between you and utter chaos!
For example:
- original – for original sequencing data (compressed FASTQ files)
- sub-directories named, for example, by year_month.<sequencing run/job or project name>
- aligned – for alignment data (BAM files, etc)
- sub-directories named, e.g., by year_month.<project_name>
- analysis – further downstream analysis
- reasonably named sub-directories, often by project
- refs – reference genomes and other annotation files used in alignment and analysis
- sub-directories for different reference genomes and aligners
- e.g. ucsc/hg38/star, ucsc/sacCer3/bwa, mirbase/v20/bowtie2
- code – for scripts and programs you and others in your organization write
- ideally maintained in a version control system such as git.
- can have separate sub-directories for people, or various shared repositories.
Download from a link – wget
Well, you don't have a desktop at TACC to "Save as" to, so what to do with a link? The wget program knows how to access web URLs such as http, https and ftp.
Get ready to run wget from the directory where you want to put the data.
Don't press Enter after the wget command – just put a space after it.
mkdir -p $SCRATCH/archive/original/2024.core_ngs cd $SCRATCH/archive/original/2024.core_ngs wget
Here are two web links:
- https://web.corral.tacc.utexas.edu/BioinformaticsResource/CoreNGS/yeast_stuff/Sample_Yeast_L005_R1.cat.fastq.gz
- https://web.corral.tacc.utexas.edu/BioinformaticsResource/CoreNGS/yeast_stuff/Sample_Yeast_L005_R2.cat.fastq.gz
Right-click (Windows) or Control+click (Mac) on the 1st link in your browser, then select "Copy link location" from the menu. Now go back to your Terminal. Put your cursor after the space following the wget command then either right-click (Windows), or Paste (Command-V on Mac, Control-V on Windows). The command line to be executed should now look like this:
wget http://web.corral.tacc.utexas.edu/BioinformaticsResource/CoreNGS/yeast_stuff/Sample_Yeast_L005_R1.cat.fastq.gz
Now press Enter to get the command going. Repeat for the 2nd link. Check that you now see the two files (ls), or tree $SCRATCH to see your Scratch directory hierarchy:
By default wget creates a file in the current directory matching the last component of the URL (e.g. Sample_Yeast_L005_R1.cat.fastq.gz here). You can change the copied file name with wget's -O option.
Also note that if you execute the same wget more than once, subsequent local files will be named with a .1, .2, etc. suffix.
Copy from a corral location - cp or rsync
Suppose you have a corral allocation or stockyard area where your organization keeps its data, and that the sequencing data has been downloaded there. You can use various Linux commands to copy the data locally from there to your $SCRATCH area.
cp
The cp command copies one or more files from a local source to a local destination. It has many options, but the most common form is:
cp [options] <source_file_1> <source_file_2> ... <destination_directory>/
Make a directory in your Scratch area and copy a single file to it. The trailing slash ( / ) on the destination says the destination is a directory.
mkdir -p $SCRATCH/data/test1 cp $CORENGS/misc/small.fq $SCRATCH/data/test1/ ls $SCRATCH/data/test1 # or.. cds mkdir -p data/test1 cd data/test1 cp $CORENGS/misc/small.fq . # or.. mkdir -p ~/scratch/data/test1 # use the symbolic link in your Home directory cd ~/scratch/data/test1 cp $CORENGS/misc/small.fq . ls
(Read more about using Absolute or Relative pathname syntax)
Now copy an entire directory to your Scratch area. The -r option says recursive.
mkdir -p $SCRATCH/data cds cd data cp -r $CORENGS/general/ general/
Exercise: What files were copied over?
local rsync
The rsync command is typically used to copy whole directories. What's great about rsync is that it only copies what has changed in the source directory. So if you regularly rsync a large directory to TACC, it may take a long time the 1st time, but the 2nd time (say after downloading more sequencing data to the source), only the new files will be copied.
rsync is a very complicated program, with many options (http://rsync.samba.org/ftp/rsync/rsync.html). However, if you use the recipe shown here for directories, it's hard to go wrong:
rsync -avW local/path/to/source_directory/ local/path/to/destination_directory/
Both the source and target directories are local (in some file system accessible directly from lonestar6). Either full or relative path syntax can be used for both. The -avW options above stand for:
- -a means "archive mode", which implies the following options (and a few others)
- -p – preserve file permissions
- -t – preserve file times
- -l – copy symbolic links as links
- -r – recursively copy sub-directories
- -v means verbose
- -W means transfer Whole file only
- Normally the rsync algorithm compares the contents of files that need to be copied and only transfers the different parts.
- For large files and binary files, figuring out what has changed (diff-ing) can take more time than just copying the whole file.
- The -W option disables file content comparisons (skips diff-ing).
- Files are only copied if their modification date is more recent or the file size is different
Since these are all single-character options, they can be combined after one option prefix dash ( - ). You could also use options -ptlrvW, separately, instead of using -a for "archive mode".
Always add a trailing slash ( / ) after directory names
The trailing slash ( / ) on the source and destination directories are very important for rsync – and for other Linux copy commands also!
rsync will create the last directory level for you, but earlier levels must already exist.
Let's copy a directory using rsync. We'll also add the -P option to show Progress as the copy progresses.
mkdir -p $SCRATCH/data cds rsync -avW -P $CORENGS/custom_tracks/ data/custom_tracks/
Exercise: What files were copied over?
Now repeat the rsync and see the difference.
Use the Up arrow to retrieve the previous command from your bash command history.
rsync -avrW /work/projects/BioITeam/projects/courses/Core_NGS_Tools/custom_tracks/ data/custom_tracks/
The bash shell has several convenient line editing features:
- use the Up arrow to scroll back through the command line history; Down arrow goes forward
- use Ctrl-a to move the cursor to the start of a line; Ctrl-e to the end
- Ctrl-k ("kill") to delete all text on your command line after the cursor
- Ctrl-y ("yank") to copy the last killed text to where the cursor is
Once the cursor is positioned where you want it:
- Just type in any additional text you want
- To delete text after the cursor, use:
- Delete key on Windows
- Function-Delete keys on Macintosh
- To delete text before the cursor, use:
- Backspace key on Windows
- Delete key on Macintosh
(Read more about Command line history and editing)
Copy from a remote computer - scp or rsync
Provided that the remote computer is running Linux and you have ssh access to it, you can use various Linux commands to copy data over a secure connection.
The good news is that once you have learned cp and local rsync, remote secure copy (scp) and remote rsync are very similar!
scp
The scp command copies one or more files from a source to a destination, where either source or destination, or both, can be a remote path.
Remote paths are similar to local paths, but have user and host information first:
user_name@full.host.name:/full/path/to/directory/or/file
– or –
user_name@full.host.name:~/path/relative/to/home/directory
Copy a single file to your $SCRATCH/data/test1 directory from the server named dragonfly.icmb.utexas.edu, using the user account corengstools. When prompted for a password, use the one we have written to the Zoom chat (or copy/paste the password from this file: $CORENGS/tacc/dragonfly_access.txt)
cat $CORENGS/tacc/dragonfly_access.txt cds mkdir -p data/test2 scp -p corengstools@dragonfly.icmb.utexas.edu:~/custom_tracks/progeria_ctcf.vcf.gz ./data/test2/ tree ./data/test2
Notes:
- The 1st time you access a new host the SSH security prompt will appear
- You will be prompted for your remote host password
- for security reasons characters will not be echoed
- The -r recursive argument works for scp also, just like for cp
- The -p argument says to preserve the file's last modification time
- otherwise the last modification time of the local copy will be when the copy was done
remote rsync
rsync can be run just like before, but using the remote-host syntax. Here we use two tricks:
- The tilde ( ~ ) at the start of the path means "relative to my Home directory"
- We use the tilde ( ~ ) in the destination to traverse the ~/scratch symbolic link created in your home directory.
cat $CORENGS/tacc/dragonfly_access.txt rsync -avrW corengstools@dragonfly.icmb.utexas.edu:~/custom_tracks/ ~/scratch/data/custom_tracks/
Exercise: Was anything copied?
Scavenger hunt exercise
Here's a fun scavenger hunt for more practice of the commands you've learned so far.
To get started:
cd cp -r /work/projects/BioITeam/projects/courses/Core_NGS_Tools/linuxpractice/what what # or using the $CORENGS environment variable cp -r $CORENGS/linuxpractice/what what cd what cat readme
Where are you when you're all done?
Welcome to the University Wiki Service! Please use your IID (yourEID@eid.utexas.edu) when prompted for your email address during login or click here to enter your EID. If you are experiencing any issues loading content on pages, please try these steps to clear your browser cache.