Close this window

Email Announcement Archive

[Users] NERSC Systems Outage

Author: David Turner <>
Date: 2011-07-06 14:27:37

Dear NERSC User, On Wednesday and Thursday, July 13 and 14, there will be scheduled maintenance on several NERSC systems, including the NERSC Global Filesystem (NGF). This email provides details of the outages. All times are Pacific Daylight Time. Wednesday, July 13 08:00 - Hopper, Carver/Magellan, Euclid, Dirac, Data Transfer and Science Gateway nodes, and /project and /global/scratch file systems removed from service 13:00 - Hopper returns to service 17:00 - Euclid, Data Transfer and Science gateway nodes, and /project and /global/scratch file systems return to service 20:00 - Carver/Magellan and Dirac return to service Thursday, July 14 08:00 - Franklin removed from service 13:00 - Franklin returns to service Advice for Hopper and Franklin Users Hopper will be missing the /project and /global/scratch file systems on Wednesday, July 13 from 13:00 until 17:00 (note that Hopper's local /scratch and /scratch2 file systems will remain available during this time). Franklin will be missing /project from 08:00 Wednesday, July 13 until 13:00 on Thursday, July 14. A batch job that requires either of these file systems will likely fail if it begins execution during the above time frames. If your job requires /project (on either Hopper or Franklin), you can protect it from failure by including the following directive in your batch script: #PBS -l gres=project:1 If your job requires /global/scratch (on Hopper), you can protect it from failure by including the following directive in your batch script: #PBS -l gres=gscratch:1 If your Hopper job requires both /project and /global/scratch: #PBS -l gres=project:1%gscratch:1 Jobs that include the above directives will not be started by the batch system during the scheduled outages of the specified file systems. This will preserve your job's priority rank in the queue. In general, if your batch jobs require /project and/or /global/scratch on Hopper, Franklin, Carver/Magellan, or Dirac, it is a good idea always to include the appropriate "gres" directive. In that way, your jobs will be able to survive scheduled file system maintenances without any further intervention. Note that Carver/Magellan (and Dirac) will be down during the entire NGF maintenance period, and so the use of gres is not necessary. -- Best regards, David Turner User Services Group email: NERSC Division phone: (510) 486-4027 Lawrence Berkeley Lab fax: (510) 486-4316 _______________________________________________ Users mailing list

Close this window