Statement on Data Loss

Special Statement Concerning Data Loss (Dec 21, 2016)

Dear Proteus Users,

About two weeks ago, the three-year-old 140-TB storage array attached to Proteus catastrophically malfunctioned. As a result, there has likely been significant data loss across the /home and /mnt/HA/groups filesystems. Several attempts to access this data show that the array is unable to provide read access at rates sufficient to provide full recovery within any reasonable time period. The UCRF board has therefore ended blanket recovery efforts. The Proteus storage array will be completely rebuilt with additional safeguards to lower risk.

The Board would like to apologize to anyone affected adversely by this data loss, and we assure all uses that the root cause of this malfunction has been identified and will be fixed. The risk of data loss of this kind was always present, given that, as a matter of policy, scientific computational data backup is the responsibility of end-user researchers. The URCF board is investigating offering an optional backup service for such data rather than relying on each research group to implement its own data backup strategy. It is hoped that such a surface will lower the barriers to protected research data.

We have entered a by-request data recovery period. Requests for specific named files or drectories should be sent to URCFS@drexel.edu by 3am on December 26. Data recovery will be attempted through 8am on January 3, 2017, but there is no guarantee that data will be recoverable or valid. (Large requests are less likely to result in successful recovery.) On January 9, 2017, all remaining data will be lost when the process to rebuild Proteus begins. Proteus should be fully operational on January 9. Questions about this information can be directed to anyone on the URCF board.

Sincerely, Cameron F. Abrams, Chair (outgoing) of the URCF Board