Peregrine System Time Updates March 2016
Here is a summary of updates applied during the recent system time:
- Lustre has been patched. This appears to have resolved the stability issues were seeing.
- Automatic purges on /scratch have been enabled. We initially set the automatic purge to delete files that have not been read in a year or more. We expect we will need to set automatic purges for less than a year though to increase free space on /scratch available for short term use. Reminder: we are working toward a 28 day purge policy. Please ensure your workflow involves copying critical output to either the /projects file system, or the /mss file system within 28 days of a file being created.
- InfiniBand repairs and updates: This involved cable replacement and firmware updates to improve InfiniBand stability. We don't expect this to affect your applications, but please send email to firstname.lastname@example.org if you observe new behavior.
- New versions of Moab, Torque, and Nitro were installed. We are seeing much faster job launch times than with the previous version. The showq command has a new filter to filter node feature strings. For example: showq -w nodefeature=64 -u USERID. For more options see the man page on showq.
- Updated node image: changes were limited to upgrading versions of scripts and software we use to manage nodes. We don't expect an impact to your applications, but please send email to email@example.com if you observe new behavior.
last modified Mar 21, 2016 02:21 PM