Backup and archiving
|Archival data handling|
|CESNET Storage Department||
This page provides information about data archiving in MetaCentrum, see Archival data handling for additional information.
What is archived
Data can be endangered by various system malfunctions or accidental or intentional means, while formulating this policy following threats were considered:
- failure of primary media (disks or disk arrays) or system software
- uncontrolled changes or destruction of data due to failure, that is not directly related to storage device (e.g. memory errors)
- destruction of data media by force majeure or external influence (e.g. fire, power surge)
- inadvertent deletion or overwriting of files by user.
Analysis of the threats enumerated above implicates need to ensure that data of permanent value is regularly backed up by establishing and following an appropriate system backup procedure. (It is not necessary to back easily recreatable data up.)
Therefore, these data are regularly backed up in MetaCentrum:
- application software installed by administrators
- user home directories, in AFS and cluster frontends
- system data and logy of servers and cluster frontends
- other data areas according to individual needs of users and user groups (common project directories etc.).
Un-backed up clusters:
- Generally, cluster work nodes (the directories for temporal data storage --
/scratch.shared) are not backed up.
Backed up clusters:
- Clusters AFS/afs are designed for long-term files storage.
- Clusters NFSv4 /storage
- Clusters /home
According to current technical circumstances (load on backup storage devices and machine being backed-up) automated backup is run approximately tree times a week.
Backups are retained for maximal time period allowed by backup storage capacity, for three months at least.
Resilience against catastrophic events is implemented by double tape libraries which are used for replicated storage of backup data. Off-site storage of backup media in not used.
Full backup on tapes is stored once a month, other backups are differential or incremental. Because files for differential and incremental backups are selected according to file timestamps, it is not guaranteed files with timestamps changed backwards (e.g. extracted from archives with original timestamps) will be copied during those backups.
A full backup of a large data set may take a long time to complete. To avoid downtime, high-availability systems may instead perform the backup on a snapshot—a read-only copy of the data set frozen at a point in time—and allow applications to continue writing to their data. Therefore the high capacity disk array of the CERIT-SC center (/storage/brno3-cerit/home...) uses snapshots for data back-up. Each snapshot is available for 14 days.
How to store your archival data
CESNET provides an infrastructure of hierarchical storages suitable for archiving purposes. Hierarchical data storage (HSM) in Jihlava and Plzen are directly accessible from all MetaCenter and CERIT-SC nodes. The storage is mounted in /storage/jihlava2-archive/home/ and /storage/plzen2-archive/home/.
MetaCentrum users obtained a space with a standard 5TB disk quota. The quota can be increased on request. Older data is moved to tapes and MAID.
The properties of the storage make it slightly differ from practices regarding MetaCentrum storage handling:
- Home directory /storage/jihlava2-archive/home/<login>/ serves to store configuration files only, it has a tiny quota and mustn't be used to keep actual user data. The actual data space is linked from the home directory and can be found under /storage/jihlava2-archive/home/<login>/VO_metacentrum-tape_tape/. Migration policy in charge ensures that the data is always kept in a redundant manner.
- All the volumes configured on the Jihlava storage facility are mounted in MetaCentrum (including service directories). If you keep your data in another virtual organisation space in Jihlava, you can access them from MetaCentrum nodes, too.
The storage facility is suitable mainly for archive data storage, i.e., data which is not accessed on regular basis. You're kindly requested not to use it for live data, especially data actively used for computations. The storage is organised in a hierarchical manner. It means the system automatically moves less used data to slower tiers (mainly magnetic tapes and MAID). The data is still available for the user in the file system. It is necessary to keep in mind that access to data unused for a long time may be slower.
There is almost no space left on Brno's disk arrays. Please consider to move your archieval data from /storage/<location>/home/ to /storage/plzen2-archive/ or /storage/jihlava2-archive/ (HSM) Moreover you get a benefit of 2 copies of your data thanks to the migration policy of the HSM.
List of MetaCentrum storages: https://wiki.metacentrum.cz/wiki/NFS4_servery Actual usage of storages: http://metavo.metacentrum.cz/pbsmon2/nodes/physical
The documentation of the directory structure can be found on https://du.cesnet.cz/wiki/doku.php/en/navody/home-migrace-plzen/start The complete storage facility documentation: https://du.cesnet.cz/wiki/doku.php/en/navody/start
The hierarchical storage is operated by the CESNET Data storage department, http://du.cesnet.cz. User support is provided by the standard MetaCenter user support firstname.lastname@example.org.