Slow access to Crex closed

There have been indications of slow I/O operations on the project storage system on Rackham and Snowy. The reason might be a problem with Crex or a heavy load to the metadata servers from I/O-intensive jobs.

Investigations will start tomorrow morning, but in the meantime we recommend users to use the local scratch storage for jobs with intensive file access. See our user guide for details - https://www.uppmax.uu.se/support/user-guides/how-to-use-the-nodes-own-hard-drive-for-analysis/

Update 2020-02-18 17:30

The problem appears intermittent and is believed to be partly related to I/O-intensive jobs. We have some indications that the upgrade to Lustre 2.12 may have changed our performance profile, which we are also investigating. The performance recovered somewhat during the afternoon but has had several spikes with severe metadata operations, such as listing files using ls or changing directories using cd. We will continue to work on this issue.

Update 2020-06-16 16:00

We have seen a few spikes of lower performance since the end of february but we do not believe there is a general issue of bad performance with Crex. Please contact the support at support@uppmax.uu.se if you have any further questions or reports of this problem.