Skip to content

Data sharing

Sometimes two or more MetaCentrum users need to work on the same files and directories.

Since setting whole home directory as accessible to anyone is considered a security vulnerability (due to the possibility to manipulate sensitive files like .k5login, .bashrc by a third party) such settings are discouraged and are automatically reset.

To make sharing of data possible and at the time safe, we offer two options varying slightly with respect to expected volume of shared data, namely:

  1. sharing based on a group (moderate amount of data)
  2. sharing based on project directory (large amount of data)

What is the difference between group and project directory?

With project directory the data are not accounted to users' quotas, therefore project directory is best suited for sharing a large volume of data and/or large amount of files which could otherwise fill individual users' quotas easily.
On the other hand, in sharing based on group the data are accounted to users quotas based on the ownership of individual files.

Group setup

  1. Choose a name for your group and ask us at meta@cesnet.cz to create it
  2. Include list of members of the group and choose at least one group manager

The group manager can later add/remove members of this group through the Perun interface.

After the group is created, you will be notified.

Finally, create a directory dedicated for shared data, e.g. /storage/CITY/home/USER/shared.

Project directory setup

  1. Choose a name for your group and ask us at meta@cesnet.cz to create it together with a project directory
  2. Include list of members of the group and choose at least one group manager

The group manager can later add/remove members of this group through the Perun interface.

After the group is created, you will be notified.

There will be also a /storage/projects/GROUP_NAME directory created to store the data in.

Using the shared directory

The usage of shared directory is the same for both shared directories in /storage/projects and in /storage/CITY/home/USER/.

  1. assign the directory to your group (denoted by "MYGROUP")

    chgrp -R MYGROUP /storage/CITY/home/USER/shared

  2. set a "sticky bit" (permission atribute) to the directory, so that all the data created within it will belong to the group

    chmod g+s /storage/CITY/home/USER/shared

  3. Ensure the correct identity within the system/job environment

When creating files, the system has to know under which group identity the files should be created.

Here, it is necessary to distinguish between the work on frontends and work within running jobs.

Identity on frontends

To create files when working on frontends, it is necessary to change your primary group to the requested one:

newgrp MYGROUP # all the created files will be owned by the MYGROUP group
umask 002 # will set the access rights of all the newly created files to rwxrwxr-x

You can include the above lines into system start-up scripts /storage/*/home/USER/.profile or equivalent file (e.g. /storage/*/home/USER/.bash_profile) so that the correct group is set automatically.

Alternatively, you can ask us to change your primary group within the whole MetaCentrum infrastructure.

Identity in running jobs

When creatingfiles within running jobs, it is necessary to have the following lines within your batch script:

#!/bin/bash
#PBS -W umask=002
#PBS -W group_list=MYGROUP
- option -W umask=002 ensures the correct access rights to created files (rwxrwxr-x) - option -W group_list=MYGROUP ensures that all the processes will run under the group MYGROUP

these options further ensure an availability of the $SCRATCHDIR directory to the group members as well - the directory will be owned by the requested group and the access rights will be rwx:

$ qsub -I -W group_list=einfra -W umask=002
qsub: waiting for job 4025666.meta-pbs.metacentrum.cz to start
qsub: job 4025666.meta-pbs.metacentrum.cz ready

konos6$ ls -ld $SCRATCHDIR
drwxrwxr-x 2 makub einfra 6 Feb 13 10:50 /scratch/makub/job_4025666.meta-pbs.metacentrum.cz

konos6$ id
uid=13153(makub) gid=10002(einfra) groups=10000(meta),10002(einfra),10100(storage)

Warning

Because of a bug in the Network File System (NFS) we use, it is necessary to explicitely change the group ownership of the newly created files/directories (at the end of an interactive session or job) by calling the command

chgrp -R MYGROUP DIRECTORY

(Otherwise, the data will be saved under your primary group.) Alternatively, you can ask us for changing your primary group throughout the whole MetaCentrum infrastructure.

sync_with_group usage

Some users find changing umask inconvenient. For these users we recommend another approach.

  1. Copy the files you need to work with from shard directory elsewhere.
  2. Process the data, create new files etc.

  3. When ready to share the data, send them back into shared directory by a command

    sync_with_group group_name source_dir target_dir

where

group_name = name of the group the project directory belongs to
source_dir = the working directory
target_dir = shared project directory