March 24, 2014
Newman Lab Network Outage: Tuesday, March 25 @ 7:30 AM
There will be brief network outage (including RedRover & eduroam wireless
) in Newman Lab, starting at 7:30 AM on Tuesday, March 25, and expected to last 30-45 minutes. The purpose of the outage is for CIT to activate a new high-speed connection between Newman and CCC, which will eventually provide 10-gigabit network connectivity among the buildings of CLASSE (currently 1-gigabit).
We try to standardize and maintain uniformity with all of our linux systems as much as possible to maximize the resources available to you, secure your data, and minimize support overhead.
For example, we try to test any new system-level packages, manage the install with our configuration management system, and eventually push them to all CLASSE linux systems.
Likewise, the only areas on the local system disk that can be written to are /tmp or /var/tmp. In general, each project or group has central NFS file systems available to them for installing software and storing data, and each user has directories for personal data and program files.
For example, a CHESS user may install software in /nfs/chess/sw/, store any data in /nfs/chess/data1/, and store any user files in /nfs/user/username/. /home/username/ should mainly be used for program files, init scripts, etc. Anything you install in these directories from your desktop will be on high-avaibility and redundant central storage, and will work on any other CLASSE Linux system (for example our compute farm / batch queuing system, our various control systems, etc.). You will also be able to access these files from Windows at \\samba\chess_data1, \\samba\user\username, and \\samba\home\username or from OS X at smb://samba/chess_data1, smb://samba/user/username, and smb://samba/home/username.
For more on Linux at CLASSE and data management, please see LinuxSupport
Linux nvidia driver update
On Tuesday 3/25, all CLASSE Linux systems with nvidia graphics cards will be updated to the latest nvidia driver to enable CUDA
CUDA availability on Linux
toolkit is now available from all CLASSE Linux systems in /nfs/opt/cuda. For more information, please submit a service request. CLASSE usage documentation is in the process of being written.
CLASSE Compute farm upgrade to support nVidia GPU/CUDA
To support the newly purchased nVidia GPU/CUDA compute node, the CLASSE Compute farm will be migrated to an upgraded Open Grid Scheduler/Open Grid Engine (OGS/OGE) server running as clustered service on the CLASSE cluster. To complete the upgrade, the current farm queues will be disabled on Monday, 31-Mar-2014, allowing running jobs to complete but no new jobs will be accepted. By Wednesday (02-Apr-2014) afternoon, the queues will be completely drained of running jobs, and the old SGE server will be decommissioned. The new OGE clustered service will then start accepting new jobs. Our Grid Engine wiki (https://wiki.classe.cornell.edu/Computing/GridEngine
) will be updated with instructions on how to instruct the queuing system to use GPU hardware.
Email address in Service Request Web form
Please be sure to provide a valid email address (e.g. NetID@cornell.edu
) when you use the Service Request form at https://wiki.classe.cornell.edu/Computing/ServiceRequest
If the address is wrong, you won't receive any acknowledgement and attempts to help you will disappear into the ether.
General network and server maintenance will occur every Tuesday from 12:00 noon to 2:00 PM.
The CLASSE-IT group will always announce any expected disruptions in our NewsLetter
and via CLASSE-IT-NEWS-L
, but with the size and complexity of our network there is always the potential for something to go wrong. We will do our best to contain all network maintenance and planned outages to Tuesdays from 12:00 noon to 2:00 PM.
Unless other arrangements have been made, CLASSE-managed Windows systems may be updated and rebooted on Tuesday morning at 2:00 AM
, so please avoid critical or lengthy operations at that time. For more details, please see SystemExpectations
Questions or problems? Submit a service request.