Important message about SCRATCH area on Marconi - UPDATE

Dear Users,

as mentioned in our previous communication, the high occupancy of the SCRATCH area caused this filesystem to become unresponsive, and its access (such as listing, reading or writing files) unsuccessful. This most likely caused the failure of all jobs which needed to access the scratch filesystem.

In order to restore the situation we had to

- unmount the scratch partition on the login nodes and all the compute nodes

- reboot Marconi login nodes.

We would kindly ask you to:

1) put in "Hold" state all currently pending jobs which need to access the scratch area with the command "scontrol hold <jobid>"; once the situation will be recovered, and the scratch area will be mounted back on the compute nodes, we will ask you to release the jobs.

2) wait until the scratch partition will be made available again on Marconi login nodes (we will inform you when done); we will then ask you to check your data and reduce the occupancy of your $CINECA_SCRATCH to the essential data you need to keep in that area, and eliminate/move elsewhere what not needed by your current activities.

We apologize for the inconvenience,

HPC User Support @ CINECA