Software error causes 34m research files to be lost at Kyoto University

5 Jan 2022

Image: © Pixza/Stock.adobe.com

A new script malfunctioned and deleted 34m files from the university’s supercomputer, some of which cannot be recovered by backups.

Around 77TB of research files have been lost at Kyoto University following a malfunction caused by a supercomputer software update.

On 16 December, the university began to investigate an issue caused by a Hewlett Packard Enterprise (HPE) software update. It found that approximately 34m files from 14 research groups had been deleted in the previous two days, with data from four of the groups unable to be recovered from backups.

The university said it would immediately contact those affected.

HPE took full responsibility and said the update for Kyoto University’s supercomputer system was originally designed to “improve visibility and readability,” by deleting log files older than 10 days.

It admitted in a statement that there was a “lack of consideration” in the release procedure of the new script and it was not aware of any potential side effects when it was applied.

This led to a script being overwritten while it was still running, “resulting in undefined variables” that caused the original log files in the supercomputer to be deleted “instead of deleting the [log] file saved in the directory”.

“We believe this file loss is 100pc our responsibility,” the company added, and linked to a support email to help those affected with file recovery and loss reduction techniques.

HPE said measures will be taken to ensure this issue does not occur in the future, including verification of updates before application and re-educating engineers in charge of human error and risk prevention.

Kyoto University is the second oldest university in Japan, founded in 1897. It is one of Japan’s leading research-oriented institutions and has produced multiple Nobel Prize winners.

Microsoft faced a software issue of its own at the start of the year, when its Exchange servers stopped working properly as the clock struck midnight on New Year’s Eve. The servers could not accommodate the year 2022, leading to some calling it the Y2K22 bug.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Leigh Mc Gowran is a journalist with Silicon Republic

editorial@siliconrepublic.com