Skip to end of metadata
Go to start of metadata


 Table of Contents - click to expand/contract

Background

Every user on Computerome has a HOME directory (/home/people/<user>), which contains for instance environment setup and everything else which is considered strictly user specific. For security reasons, HOME directory permissions are set (and enforced) so that only the user has access.

$ ls -ld /home/people/someuser
drwx------ 5 someuser someuser 235 Jun 24 10:17 /home/people/someuser

Because users will sometimes switch projects, graduate, change employment, etc., no project data or anything else project related should ever be kept in the HOME directory, but must be maintained in the Projects structure.

Projects in Computerome are identified by the following information:

  • <PROJECT>
    How the project is known to the world; usually also funding the project.
  • <project_NAME>
    How the project is referenced in DTU and Computerome; also the top-level name for the projects directory structure
  • <group_NAME>
    Official id assigned for the project by DTU Basen; this <group_NAME> is also the UNIX group, that controls the projects access to resources.Each individual <user> must be a member of Unix group <group_NAME> to be able to access data and submit jobs in the project.Individual <user>'s are added to or removed from Unix group <group_NAME>, depending on whether they should work in the project or not.

It is recommended, that <project_NAME> corresponds to the <PROJECT> that is funding the project.

In most cases (but not always) <group_NAME> and <project_NAME> will be the same, and have a format similar to pr_xxxxx or ku_xxxxx; for example pr_12345.
The simplest way to make sure is to first check your group memberships with the id command, and look for group(s) that do not match uid/gid:

$ id
uid=12321(lifesci) gid=12321(lifesci) groups=12321(lifesci),2345(pr_12345),4321(pr_54321)

and then look for those groups in the /home/project/ directories:

$ ls -l /home/project | grep -e pr_12345 -e pr_54321
drwxr-x---   8 pamar   pr_12345   147 Jan 19 12:45 pr_12345
drwxr-x---   4 pawbak  pr_54321    48 Sep 26 12:27 blip

To give <user> access to cluster data, batch jobs must be submitted with parameters '-W group_list=<group_NAME>' and '-A <group_NAME>' , where '-A' supplies account information for use in Moab Accounting Manager (MAM).



Project file structure

New projects are created with the following default directory structure:

/home/projects/<project>
./apps
./apps/modulefiles
./archive
./data
./data/generated
./data/raw
./people/<user>
./review
./scratch

The recommended usage is as follows:

  • /home/projects/<project>
    projects HOME directory
  • ./apps
    project specific applications - used when the project, for whichever reason, can not use the standard application provided in for instance /services/tools.(Common candidates include: anaconda, perl, qiime, R, ncbi-blast, samtools, bamtools, bedtools, java)
  • ./apps/modulefiles
    project specific modulefiles./archive./dataproject specific data
  • ./data/raw
    raw data
  • ./data/generated
    generated data
  • ./people/<user>
    each project members private, project related stuff (data, scripts, etc.)
  • ./scratch
    temporary or easily re-creatable data - using this may significantly decrease size and improve speed of backups.
    NOTE: Will not be backed up and will not be archived with project.
  • ./review
    (sub)sets of data, used for peer-review of articles.

This directory structure has been selected to ease close-down and archiving at the end of a project.



Database structure

Computerome also supplies access to a series of read-only reference databases.

/home/databases