Purdue has a few policies concerning the use of the site's resources. These policies are intended to provide reasonably fair access to users of our resources, and indicate the level of service we can provide. Users who violate the policies listed below may find their access curtailed as is deemed appropriate.
- Purdue resources are primarily intended for use with the US-CMS VO at this time. If you would like to use Purdue's resources for another VO, please contact our site admins at email@example.com .
- Any planned event that would impact the regular operation of
processing at our site will by preceeded by public notices, both on
Purdue's Tier2 web site and to the USCMS Support Center, at least 96
hours in advance.
- Unplanned outages will be acknowledged on the web site and to the USCMS Support Center within the first 6 hours of downtime, and again when normal operation resumes.
- All access to processing resources is opportunistic unless otherwise arranged.
- General job policies:
- All jobs are limited to a maximum of 30 days of runtime.
- In extreme cases, we may be forced to remove running jobs. Should this happen, we will attempt to requeue submitted work, rather than remove it entirely.
- If we need to affect running jobs, we will notify the affected VO. While not all system events may allow it, we will attempt to provide up to four hours advance notice.
- Using dedicated and non-dedicated resources:
- Dedicated resources are available to CMS through the standard PBS jobmanager on lepton.rcac.purdue.edu (lepton.rcac.purdue.edu/jobmanager-pbs)
- Non-dedicated resources are available via PBS through the condor jobmanager on the Purdue-ITaP site (ie, osg.rcac.purdue.edu/jobmanager-condor) Non-CMS users or CMS users who need more than CMS's dedicated share should use Condor.
- No guarantees are offered for runtime on non-dedicated resources.
- Non-dedicated resources are generally dedicated to other local users. Jobs submitted to these systems may be preempted with no notice, and requeued.
- Preempted jobs will eventually be restarted. Any previously executed work will be lost, and the job will be completely re-run.
- Remote access to Purdue storage should primarily use either Hadoop or SRM.
- All Hadoop-accessible storage will not be backed-up.
- We may, at our discretion, remove files older than 31 days.
- All files residing in $HOME older than a week will be purged.
- If shared filesystems fill, Purdue will begin contacting the largest users. We will attempt to provide 4 hours notice before deleting files.
- Certain users or VOs may be given access to site-specific NFS-mounted storage. This space will be regularly backed-up to tape, on a backup rotation of at least 90 days.