General Backup Policy

Of all data that is backed up the so-called retention period is 30 days. This means that if the user deletes that data, the data can be restored up to 30 days after it has been deleted.

Apart from this the number of versions of a file that is backed up is 20. This implies that when a file is backed up daily and a user wants to have the version of the file as it was 19 days ago, then this version can be restored.

The above holds for the backup policies described below unless explicitly stated otherwise.

Life Science Grid

The Storage Elements of all Life Science clusters are NOT backed up. The home file systems of the User Interface machines are backed up on a daily basis. NB: see also decommissioning of the Life Science Grid.

dCache storage

When it comes to data storage there are three options for the SURF front-end storage:

  • Data is stored on disk only. For disk-only data there is no backup. Users are advised if they cannot afford to loose this data to copy it elsewhere as well.
  • Data is stored on disk with backup on tape. For data on disk and tape there will be a copy of the data both on disk and on tape all the time. When data is removed from disk, the tape copy will be removed as well after a retention period of 10 days; In contrast to what is stated in the section "General Backup Policy", here there is only one version of the file stored on tape so the retrieval of previous versions of a file will not be possible.
  • Data is stored in an archive. The data is permanently stored on tape and may be cached on disk if users need to use the data. The data that is cached on disk is prone to be garbage collected when the disk space runs out. So for this storage class there will be a permanent copy of the file on tape only. When the data is removed from the archive the data will be removed from tape after a retention period of 10 days.

These storage classes are configured for the users on the users’ request.

For all data on tape written through dCache by default two tape copies are stored at two geographically distributed sites. But at the users’ request it can be configured to be a single copy.

When a user removes data that is on tape, it will still be recoverable for 10 days.

dCache databases that are used to store metadata that is needed to retrieve files from the disk pools are backed up on a daily basis. In addition, warm fail-over is possible through the mechanism of so-called point-in-time recovery.

Machines where users have an interactive login

For these machines the users home directory is backed up on a daily basis. An example of such a node is the Grid user interface machine ui.grid.sara.nl.

Machines that run database services for users

The databases on these nodes are backed up on a daily basis.

Machines that run services for specific user communities

The machines that run services for specific user communities or users are backed up on users request.

Topos

The topos service is backed up daily.

EPIC

Our persistent identifier services are backed up daily.

iRODs

The iRODs service is backed up daily.

Specific Grid Services

  • UI is backed up daily.
  • WMS is not backed up. Output of jobs may be lost in case of problems with the underlying storage system.
  • LFC, FTS is backed up. The database on these nodes is backed up on a daily basis.
  • BDII is not backed up. This service does not store user data and has been deployed in a fail-over configuration.
  • VOMS has a warm standby and databases are backed up daily.
  • PX has a warm standby and is not backed up. This service does not store any user data except users’ x509 proxies.