Scratch Purge Policy ChangeUPDATE: August 29, 2017 3:13pm
Reminder, this policy change took effect today. While scratch purges announced yesterday and today will be of files unread for 90 days, the purges next week will be for all files unread in 60 days or more.
ORIGINAL: June 6, 2017 4:37pm
After a long review, Research Computing has determined it is necessary to alter the scratch storage purge policy on all systems.
Effective August 28, 2017, all scratch storage systems will begin purging files which have not been accessed (for either reading or writing) in 60 days or more on a weekly basis.
This is a change from the previous policy of 90 days. This change is necessary in order to maintain a minimal level of free space on our scratch storage systems and ensure they remain effective for everyone.
Any files which are in active use by jobs on the clusters will not be purged, nor will you need to take any steps to ensure they are not. Please refrain from explicitly taking actions to exempt files from purging on a long-term basis, as this impedes our ability to offer the generous spaces we do for all users. It may also endanger your research, because there are no backups or recovery options for lost files in scratch storage spaces.
As soon as you are no longer actively computing against data, we strongly recommend copying that data to either a Data Depot space or the Fortress Archive, where your data will be safe for any future needs. There is information on how to access both of these systems on our website, and you may contact us at firstname.lastname@example.org at any time with questions. If you need personal assistance, our staff also hold Coffee Break Consultation .