Lustre file system Hebe

go_back.gif Lustre

Overview

Access lustre.hpc.gsi.de
Path /lustre
Capacity 32.6 PiB
Aggr. Bandwidth 675 Gb/s

go_forward.gif Usage Statistics

go_forward.gif Lustre Throughput Monitoring

UPDATED on 2021-05-17

Data safety

Warning! GSI Lustre storage is not backed up!

Ie. if you accidentally delete a file it cannot be restored.

The risk of system-side data loss is minimized by redundant storage on RAID arrays that can tolerate the simultanous failure of multiple disks. However, user or administrator errors or technical failures can end up in data loss.

We take measures to be able to restore the file system even in case of desaster but likely the data will be outdated and/or incomplete.

Unified Lustre mount point

  • /lustre is the path to access the Lustre file system(s).
  • The paths /lustre/nyx and /lustre/hebe remain as a symbolic links for backwards compatibility
  • The underlying Lustre file system for the path /lustre as of December 2019 is the Hebe Lustre file system.

Quotas

Two distinct quotas limit the usage on Hebe:
  1. a limit on the maximum occupied space
  2. a limit on the number of files

Storage space quota

The storage space is limited on a per-group basis. When a group exceeds its assigned quota, warning mails are sent to the associated group coordinators.

If you are a member of multiple groups the group id (gid) information of the files control for which group they are accounted for. The gid info can be changed with the chgrp command.

The following command will assign the group mygroup to all your files beneath /lustre/mygroup/mydir
ยป chgrp -Rc mygroup /lustre/mygroup/mydir

Group quotas are negotiated between the group coordinators and IT in the course of the Computing-Users-Meeting.

Quota on the number of files

In addition to the storage quota we limit the number of files (inodes). This limit is implemented as the number of files has a signifikant impact on the overall performance of the Lustre filesystem.

Each file system object (directory, file) uses at least one inode. Large files and directories may require multiple inodes.

The limitation is implementet per-user. Warning mails are sent out when the limit is exceeded. The default limit is 300 000 inodes.

Individual limits may be increased if a conclusive rationale is provided eg. for project accounts. To apply for a raise send an email to lustre-service@gsi.de.

This topic: Linux > WebHome > FileSystems > LustreFs > HebeFs
Topic revision: 2021-05-17, ChristopherHuhn
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding GSI Wiki? Send feedback | Legal notice | Privacy Policy (german)