Problem with Lustre /scratch

We are currently experiencing a problem with the Lustre /scratch filesystem. This problem is due to a file system error on OSS-0-1. We are investigating the situation but do not yet have a time frame for resolution.

This issue is causing performance degradation and can result in the inability to write to files that were stored on the faulty portion of the file system. Depending on your configuration new writes may also fail in the current state. All reads should be completing successfully.

If you are experiencing errors that you believe are related to this problem please contact hpc-sysadmins@iowa.uiowa.edu

Update:

 
The storage server volume that was problematic has been remounted and went through a recovery process. It is now possible to write to it. However, there have been a lot of client eviction messages so I would imagine that there are issues with job output. Also, the scratch file system continues to be under an extremely heavy load.