Examples not working?

File Sizes and Disk Quotas #

File sizes on /wynton/ #

The /wynton/ storage is on a ZFS file system on top of our BeeGFS parallel storage system. This is automatically compressed (using lz4 compression in ZFS) before anything is written to the physical drives. Because of this, a 1.0 MiB file is likely to occupy less that 1.0 MiB of drive space. Exactly, how much a file is compressed varies greatly with file format but as a rule of thumb plain text files can be compressed more than files in a binary format. Already compressed files such as GZ or ZIP files are unlikely to be compressed further.

Because of this underlying disk compression, command-line tools such as ls and du may not report what you expect it to report. For example, consider an Apptainer image file rocker_r-base.sif of size 274,538,527 bytes (= 274,538,527/1024^2 = 261.8 MiB);

[alice@dev2 ~]$ ls -l rocker_r-base.sif
-rwxr-xr-x. 1 alice boblab 274538527 May  8  2018 rocker_r-base.sif

The actual space consumed on disk by this file is 256,136,704 bytes (93.3%):

[alice@dev2 ~]$ ls -s --block-size 1 rocker_r-base.sif
256136704 rocker_r-base.sif

Using the disk-usage tool du, we can see the same if we do:

[alice@dev2 ~]$ du --apparent-size --block-size=1 rocker_r-base.sif
274538527       rocker_r-base.sif

and

[alice@dev2 ~]$ du --block-size=1 rocker_r-base.sif
256136704       rocker_r-base.sif

Comment: It is the compressed size that counts towards your disk quota.

Disk quota #

The BeeGFS file system keeps track on how much disk each of us currently consumes in different storage locations, specifically:

  1. User home folder ($HOME, i.e. /wynton/home/ or /wynton/protected/home/)

  2. Group folder under (i.e. /wynton/group/, /wynton/protected/group/, and /wynton/protected/project/)

  3. User files and folders under /wynton/scratch/ and /wynton/protected/scratch/ (unlimited quota)

These different type of locations are formally referred to as storage pools by BeeGFS.

If we ever run out of quota in one storage pool, BeeGFS detects this and prevent us from writing additional data. The symptoms of a full quota may vary. You might get a clear “disk full” error, or you might experience obscure issues such as having problems logging in. You can use the BeeGFS tool beegfs-ctl --getquota ... to check how much disk quota you have and how much you currently consume. Below sections give instructions how to check this for the different storage locations available.

User disk quota on /wynton/home/ or /wynton/protected/home/ #

To check how much storage space you have consumed on /wynton/home/ or /wynton/protected/home/, and the total amount available to you, call:

beegfs-ctl --getquota --storagepoolid=11 --uid "$USER"

For example,

[alice@dev2 ~]$ beegfs-ctl --getquota --storagepoolid=11 --uid "$USER"
      user/group     ||           size          ||    chunk files    
     name     |  id  ||    used    |    hard    ||  used   |  hard   
--------------|------||------------|------------||---------|---------
         alice| 99002||   88.71 GiB| 1000.00 GiB||   645266|unlimited

tells us that user alice has 645,266 files that occupy 88.71 GiB (‘size used’) on the BeeGFS file system out of their 1000.00 GiB (‘size hard’). Importantly, because the /wynton/home/ storage is mirrored, the disk usage (‘size used’) and the available quota (‘size hard’) are reported at twice the size of what you would expect for a non-mirrored storage. This is why your 500-GiB home storage space is reported as 1000 GiB by the beegfs-ctl tool.

Group disk quota on /wynton/group/, /wynton/protected/group/ and /wynton/protected/project/ #

If your group/lab (e.g. boblab) has purchased additional storage, it is available under /wynton/group/, and possibly also under /wynton/protected/group/ (PHI), or /wynton/protected/project/ (PHI). To check how much storage space your group/lab has consumed of the total amount available to it, call:

beegfs-ctl --getquota --storagepoolid=12 --gid "$(id --group)"

For example,

[alice@dev2 ~]$ beegfs-ctl --getquota --storagepoolid=12 --gid "$(id --group)"
      user/group     ||           size          ||    chunk files
     name     |  id  ||    used    |    hard    ||  used   |  hard
--------------|------||------------|------------||---------|---------
        boblab| 34001||   13.43 TiB|   40.00 TiB||        0|unlimited

says that the boblab group is using 13.43 TiB out of the 40.00 TiB group storage they have acquired. If the hard limit is 1 Byte (sic!), that is a placeholder that the group does not have purchased any group storage.

The group storage is shared among all group members and does not count toward your personal disk quota under $HOME.

Any group with purchased storage can have a group folder in both the non-PHI /wynton/group/ area and the PHI /wynton/protected/group/ area at the same time, e.g. /wynton/group/boblab/ and /wynton/protected/group/boblab/. In that case, the group quota usage would include group-owned files in both areas. Your lab can also request to create a subgroup (e.g. boblab-phi) to self-manage the quota for PHI-only data storage (e.g. /wynton/protected/group/boblab-phi/), and it could be allocated as part of the parent quota.

PHI projects under /wynton/protected/projects/ are for controlling access to IRB related data, where the IRB access group does not encompass the whole PI group or where the access group encompasses selected members of more than one PI group. The quota for the directory in /wynton/protected/projects/ could either be a separate group quota purchase or a portion of a purchased group quota dedicated to the project (subgroup).

Additional group storage via other groups #

It could be that you are part of another lab or institute that have purchased their own storage. This is common for users affiliated with Gladstone Institutes. For example, above, user alice is part of the boblab group;

[alice@dev2 ~]$ id --group
boblab

but they are also part of the carol_inst group;

[alice@dev2 ~]$ id --groups
boblab carol_inst

To see the disk quota for all groups, we can call:

for group in $(id --groups); do beegfs-ctl --getquota --storagepoolid=12 --gid "$group"; done

For example,

[alice@dev2 ~]$ for group in $(id --groups); do beegfs-ctl --getquota --storagepoolid=12 --gid "$group"; done
      user/group     ||           size          ||    chunk files
     name     |  id  ||    used    |    hard    ||  used   |  hard
--------------|------||------------|------------||---------|---------
        boblab| 34001||   13.43 TiB|   40.00 TiB||        0|unlimited
      user/group     ||           size          ||    chunk files    
     name     |  id  ||    used    |    hard    ||  used   |  hard   
--------------|------||------------|------------||---------|---------
    carol_inst| 35000||   75.16 TiB|  100.00 TiB|| 14957560|unlimited

which shows that alice has access to 40 TiB via the boblab group and another 100 TiB via the carol_inst group.

User disk usage on /wynton/scratch/ and /wynton/protected/scratch/ #

To check your disk consumption on the global scratch space (/wynton/scratch/ and /wynton/protected/scratch/), use:

beegfs-ctl --getquota --storagepoolid=10 --uid "$USER"

Comment: There are no user or group quotas on /wynton/scratch/ and /wynton/protected/scratch/, but files on the global scratch that are older than two weeks are deleted automatically.