HPC Backup services¶

If you are using IBM Spectrum Protect client to backup your data to HPC, here is a quick guide on the command-line usage.

The backup-client also supports a Windows based version, however this will currently not be covered in this guide.

Requesting backup services¶

While tape-backup is the main backup strategy in HPC, it usually makes sense in specific environments with specific restore needs.

If you'd like to request backup services, please contact support@hpc.ut.ee and a conversation with more specifics can be opened. The main option for tape-based backups is IBM Spectrum Protect, but other options such as rsync, NFS, SMB, Object are also available, depending on your needs.

Spectrum Protect Linux CLI¶

Using the backup client¶

Almost all of the client-side backup actions are done using the binary dsmc, which is quite primitive and simple to use as most actions are treated as sub commands. The syntax also has cisco-style shortening, meaning the minimum amount of letters from the beginning are needed(ba, bac work for backup, i, in,inc for incremental etc.)

There are many client-side options with full documentation available at the official IBM documentation website however, this guide will note some of the most important ones for daily use.

Warning

Do note the trailing slash at the end of paths, this is important since the folder is treated as a (stub)file otherwise. In general when working with folders, always append a trailing slash

Full backups¶

For full backup copies or for the first run in an incremental backup, use full

dsmc full /<path>/

By default dsmc will not descend into subdirectories if a folder is specified as the argument. For this, use the command-line argument -subdir=yes

Incremental backups¶

Incremental backups are the meat-of-the menu for most backup use cases, comparing file-differentials and only creating copies of new files or modifications.

The amount of incremental copies and retention parameters are set server-side, for more information please contact support. These can be set on a per-client basis.

dsmc incremental -su=yes /<path>/

Querying backups¶

To query backup files, issue query backup. Usually, you want to query for specific files and folders, in which case subdir might not be necessary.

query backup also supports -filelist as input.

-inactive will also show "inactive" copies and not only the latest incremental version.

dsmc query backup -ina -su=yes /<path>/

Restoring data¶

Data restores are done by path, can be both recursive or single files. Use -latest for latest incremental backup versions. If destination path is not set, files will be attempted to be restored into their original paths.

-su=yes is optional if restoring single files.

dsmc restore -latest -su=yes /<source path>/ [<destination path>]/

Archive client¶

Warning

While Spectrum Protect supports archiving, the default archive parameters in the server keep a 365 day retention. For long term archival options, please contact HPC support.

Setting up the backup-client¶

A configuration file and the binaries will be provided to you by the HPC team, along with preliminary setup instructions. However, there are a few considerations

Encrypting the backup¶

During the first initialization of the backup-client, you can also specify some options to make sure the backup is encrypted with a key stored in the local machine. The extra configuration will look similar

DOMAIN                          /backup
include.encrypt                 /backup/data
include.encrypt                 /backup/home
ENCRYPTKey                      save
ENCRYPTIONType                  AES128

During the first initialization, a password will be prompted for the encryption key.

Save this key in a secure location as the backup client will not prompt for it again unless the configuration is restored