Accessing Swestore with the ARC client

< Getting started with SweGrid
< Swestore

This guide describes how to use the Nordugrid ARC client for storing and retrieving files from Swestore. The ARC client is usually used for sending grid jobs to grid clusters, but it also contains commands for data management. A complete user guide for the ARC client can be found in http://www.nordugrid.org/documents/arc-ui.pdf.

Requirements

To access Swestore using the ARC client you need to get a grid certificate and a storage project.

All SNIC HPC systems should have the ARC client installed. If yours doesn't, please contact support at your centre so they can fix this as soon as possible. To install the ARC client on your own computer, please follow instructions here, or see the official Nordugrid ARC installation page for more information.

Quickstart

Basic commands

arcproxy - unlock your certificate so you can use it. See Proxy certificates for details.

arcls - for listing files. Works similarly to ls. Example arcls gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME

arcmkdir - for creating directories. Works similarly to mkdir. Example arcmkdir gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/newdir

arccp - for copying files. Works similarly to cp. Example arccp myfile.txt gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/myfile.txt

arcrm - for deleting files. Works similarly to rm. Example arcrm gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/whoops.txt

Use man and --help to get more info on each command. Examples: man arcrm or arcls --help

Paths

The ARC commands supports multiple storage protocols, we recommend using GridFTP with paths on the form gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/....

Copying files

Copying files to and from resources is accomplished using the arccp command.

Copying single files

Copying single files is accomplished in the same way as using the normal cp command as shown in the following example:

$ arccp archive.tar.gz gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/

Please note the trailing / which marks the destination as a directory. Without a / the destination will be a file, which may or may not be what you wanted. All required directories are created when needed so the destination may be a nonexisting directory.

Recursive copying

Recursive copying is accomplished using the --recursive option to arccp. The argument to the option determines the depth of the recursive copy, just supply a really big number like 999 if you want the entire source directory tree.

Example:

$ arccp --recursive=999 foobar/ gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/

NOTE: The above example will copy all files in the directory foobar into the destination directory YOUR_PROJECT_NAME. If you want the directory foobar to be part of the destination path you have to explicitly supply it as shown in the example below:

$ arccp --recursive=999 foobar/ gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/foobar/

Long-running operations

Note that copying large directory trees can take quite some time, and will fail if you're not aware of the following:

• Your login session created with the arcproxy command has a limited lifetime. Use arcproxy -I to show the remaining time. Use arcproxy -c validityPeriod=xxH to initiate a session with longer lifetime.
• If you loose connectivity with the resource you're running arccp on the command will abort. A utility such as screen or tmux can be used to create a terminal session you can reattach to.
• Transfer rates are largely dependent on the average file size, if you have a lot of small files the transfer will be slower than if you have large files.
• We recommend to limit your transfer sessions (ie. the directory tree copied with each arccp command) to 1TB if you have mostly large (100+MB) files and to 100GB if you have smaller files.

Listing files

Listing files on a resources is done using the arcls command. In the simplest form the command just takes a URL as input and displays names and directories without any extra information as shown in the following example:

$ arcls gsiftp://gsiftp.swestore.se/snic/bils/db/uniprot/2012_05
reldate.txt
speclist.txt
uniprot_sprot.dat.gz
uniprot_sprot.fasta.gz
uniprot_trembl.dat.gz
uniprot_trembl.fasta.gz

Additional information can be listed by adding the --long option:

$ arcls --long gsiftp://gsiftp.swestore.se/snic/bils/db/uniprot/2012_05
<Name> <Type> <Size> <Creation> <Validity> <CheckSum> <Latency>
reldate.txt file 151 2012-05-23 03:00:19 (n/a) adler32:f3f52f1d (n/a)
speclist.txt file 1715169 2012-05-23 03:00:17 (n/a) adler32:91e59dae (n/a)
uniprot_sprot.dat.gz file 462895141 2012-05-23 02:57:18 (n/a) adler32:0f131bb2 (n/a)
uniprot_sprot.fasta.gz file 79935897 2012-05-23 03:00:20 (n/a) adler32:89844c57 (n/a)
uniprot_trembl.dat.gz file 9162678278 2012-05-23 02:52:01 (n/a) adler32:b2d7cfd5 (n/a)
uniprot_trembl.fasta.gz file 4456514443 2012-05-23 02:57:34 (n/a) adler32:2b73b2a1 (n/a)

Metadata

Metadatainformation on a specific file can be listed by specifying the -m or --metadata option. Worth noting is that the amount of metadata available differs depending on which protocol is used.

Examples:

$ arcls --metadata gsiftp://gsiftp.swestore.se/ops/nikke/smallfile
/ops/nikke/smallfile
checksum:adler32:762606eb
mtime:2013-04-12 11:06:56
path:/ops/nikke/smallfile
size:30
type:file

$ arcls --metadata srm://srm.swegrid.se/ops/nikke/smallfile
/ops/nikke/smallfile
accessperm:rw-r-----
checksum:adler32:762606eb
ctime:2013-04-12 11:06:56
filestoragetype:PERMANENT
group:25001
latency:ONLINE
lifetimeassigned:PT1S
lifetimeleft:PT1S
mtime:2013-04-12 11:06:56
owner:25001
path:/ops/nikke/smallfile
size:30
spacetokens:
type:file

Creating directories

$ arcmkdir gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/newdir

If the arcmkdir command is missing the ARC utilities need to be upgraded. You can work around this by copying a dummy file to the path you want and then deleting the dummy file.

Removing files or directories

$ arcrm gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/newdir/dummyfile
$ arcrm gsiftp://gsiftp.swestore.se/snic/YOUR_PROJECT_NAME/newdir/

To remove directories they have to be empty.