CHPC News Announcements
New general KP nodes and New Guest Access Options
Posted: April 21, 2014
Kingspeak general CHPC nodes now include four 20-core nodes along with the original 32 16-core nodes. This results in new ways to request general nodes. These have been added to the “Resource Specification Section” of the Kingspeak User Guide found at https://wiki.chpc.utah.edu/display/DOCS/Kingspeak+User+Guide
Also, there are new options for guest access to compute nodes. All guest jobs are pre-emptible by any jobs submitted by the owner of the nodes and jobs using this guest access should not use /scratch/local as there is no mechanism to recover output or clean up scratch files for jobs that are preempted. Jobs run in this manner are charged to the guest account and can be run even if you still have your own allocation. Here are the available guest access options:
- For owner nodes on kingspeak and ember use ‘#PBS –A owner-guest’
- For ash.chpc.utah.edu use ‘#PBS –A smithp-guest’ (use ash-guest.chpc.utah.edu to access general ash interactive nodes)
- For telluride.chpc.utah.edu use ‘#PBS –A cheatham-guest’ (please note that this cluster is running RH5)
Finally, I would like to remind users about CHPC policy regarding interactive node usage. Any processes on the cluster interactive nodes that run longer than 15 minutes OR that impact the ability of others to make use of the node will be killed. The complete policy can be found at https://wiki.chpc.utah.edu/display/policy/2.1.1+Cluster+Interactive+Node+Policy
As always, if you have any questions regarding CHPC resources and policies, please contact us at email@example.com
Virtual School of Computational Science and Engineering this summer
Posted: April 9, 2014
CHPC will again be a sattelite site for this year's summer classes of the Virtual School of Computational Science and Engineering.
There will be two classes, each of which will take 4 days. They will both take place in the INSCC 110 Auditorium. For more information about these classes and registration please follow the links below. When registering choose University of Utah as a site.
Harness the Power of GPUs: Introduction to GPGPU Programming
June 16 – June 20, 2014
9:00 a.m. - 2:00 p.m. MST
Register at https://www.xsede.org/web/xup/course-calendar/-/training-user/class/264.
Data Intensive Summer School
June 30 – July 2, 2014
9:00 a.m. - 3:00 p.m. MST
Register at https://www.xsede.org/web/xup/course-calendar/-/training-user/class/263.
Note that these classes are open to anyone, not just to the University of Utah affiliates. We encourage participants from other institutions to register.
Matlab updated to R2014a
Posted: April 7, 2014
We have updated Matlab to the latest release, R2014a. As a part of the process we now run a new redundant license server setup which should make the license availability more resilient.
An important new feature in this release is a removal of the limit on number of parallel workers per node (previously at 12), which should be beneficial on CHPC cluster nodes with higher core counts.
For complete list of release notes for R2014a see http://www.mathworks.com/help/relnotes/index.html.
Also please note that this version of Matlab does not run on RHEL5 Linux OS, which we still run on some machines (meteo, atmos and turretarch nodes). If you use these machines, keep using R2013a by running "source /uufs/chpc.utah.edu/sys/pkg/matlab/R2013a/etc/matlab.csh" after login to that machine. OS upgrades on these machines are planned in the near future.
Guest access to ash.chpc.utah.edu now available
Posted: March 27, 2014
As was earlier announce, the smithp ember nodes were removed from ember and used to create a new cluster called ash. We are now opening up guest access to this cluster. All jobs run in this manner are pre-emptible and there are no charges against your group allocation for this usage.
Some important information:
- You need to first get the new login files from here for tcsh and from here for bash
- SSH hostids have changed. You should deleted your .ssh/known_hosts or edited this file to remove all entries for the compute nodes (emXXX)
- The chpc general interactive nodes for ash are ash5.chpc.utah.edu and ash6.chpc.utah.edu. You can use ash-guest.chpc.utah.edu and will obtain one of these two nodes. The name ash.chpc.utah.edu are the interactive nodes restricted to the Smith group.
- The cluster specific applications are in /uufs/ash.peaks/sys/pkg. We have tested what we can and all seems fine. /uufs/chpc.utah.edu/sys/pkg is also mounted as it was on ember for non-cluster specific application builds.
- If you are running your own codes, you should recompile using the new ash version of mpi, etc.
- Scratch file systems mounted are /scratch/kingspeak/serial and/scratch/ibrix/chp_gen
- It is best if you do not use /scratch/local as a guest as you do not have access to retrieve/clean up any files from this location in the likely event that your job is preempted.
- The account to use in the #PBS -A line are smithp-guest
Please address any questions and report any issues to issues
Machine Learning and Parallel Computing with MATLAB
Posted: March 26, 2014
Please join MathWorks at complimentary MATLAB seminar for educators, academic researchers and students at
University of Utah:
Date: Wednesday, April 2nd
Location University of Utah, Intermountain Network Scientific (INSCC), Room 110
Time: 9:00A.M. - 12:00P.M.
The event features one technical session presented by a MathWorks engineer:
9:00A.M. – 10:45A.M. Machine Learning with MATLAB
Learn about several machine learning techniques available in MATLAB and how to quickly explore your data, evaluate machine learning algorithms, compare the results, and apply the best technique to your problem.
11:00A.M. – 12:00P.M. Parallel Computing with MATLAB
Learn how to solve computationally and data-intensive problems using multicore processors and computer clusters. We will introduce you to high-level programming constructs that allow you to parallelize MATLAB applications without low level MPI programming and run them on multiple processors.
View complete session descriptions and register at https://www.mathworks.com/compan y/events/seminars/seminar90093.html.
telluride and ember back online
Posted: February 25, 2014
Both ember and telluride are back online and scheduling jobs. As a reminder, the smithp nodes have been split from ember, therefore the smithp-guest account will not run. Later this week or early next an announcement will be made about smithp-guest access to these nodes as part of the new ash cluster.
If you see any issues please report them to firstname.lastname@example.org.
Kingspeak, Meteo nodes, and UCS (turretarch) back in service
Posted: February 25, 2014
The above mentioned resources are now back in service. Apexarch should be up in the next hour. There was an issue with a mislabeled breaker that lead to this outage. Now that this has been resolved, the scheduled electrical work at the DDC is progressing.
KIngspeak and Apexarch Down as of approximately 8:45AM
Posted: February 25, 2014
When the electricians were shutting off the power to ember and telluride they also unexpectedly shut off the power to kingspeak and apexarch (poet) --- we are right now trying to figure out why and if the power to these clusters need to be down for the electrical work being done today. Will update as more information is available.
CHPC searching for new Director
Posted: February 25, 2014
The search for a new Director for CHPC has begun. Please see job posting PRN05344B for more details.
Emergency Network Outage -- Feb 13, 2014 starting at midnight
Posted: February 13, 2014
Duration: Midnight to 5 am
Connectivity between the Downtown Data Center and INSCC will be lost for an approximately 15-30 minute window overnight, sometime between Midnight and 5 am, to allow UEN to make networking changes. This will affect any access to CHPC file systems, including CIFS/Samba mounted shares and I/O of batch jobs. Jobs that are already running should ride through the outage. However, any batch job that attempts to start during the outage window will die when it cannot mount the file system.
We apologize for the very late notice on this outage; we put an alternate network link in place during the downtime February 6, but there were issues with the fiber and we were not able to get it repaired in time for this outage.
CHPC Downtime - affects ember, telluride and UCS clusters only - Feb 25 starting at 7am
Posted: February 13, 2014
As tentatively announced last week, CHPC will be having a downtime on Feb 25th that will require us to power down the ember and telluride clusters to allow for electrical work at the Downtown Data Center. Reservations are in place to drain the batch queues on ember and telluride by 7am on Feb 25th. The duration of the downtime undetermined at this time; we hope that the clusters will be back in service on the 25th, but there is a chance that the time frame of the electrical work will require the clusters to be down until sometime on Feb 26th.
In addition, access to the UCS (turretarch nodes) will be unavailable for part of the day as part of the final migration of equipment from the Komas Data Center.
During this downtime, CHPC will also start the separation of the smithp nodes from ember into a new cluster named ash. This separation is necessary, due to limitations of the infiniband infrastructure, in preparation for a planned addition of nodes. Once the new ash cluster is up, most likely later during the week of the 25th, guest access will be allowed as before. Details on how to access will be sent to all after the ash cluster is online.
Allocation Requests for Spring 2014 Quarter Due March 1st
Posted: February 8, 2014
This is a reminder that proposals for allocation in the general allocation pool (for use on ember and kingspeak) for the Spring 2014 quarter which starts April 1, 2014 are due by March 1, 2014. We must have this information if you wish to be considered for an allocation of time for the Spring 2014 calendar quarter. Your request may be for up to 4 quarters.
Information on the allocation process including links to the relevant forms are located online at this link.
Please let us know if you have any problems with or questions about the allocation process, or if you have suggestions on how to improve this process. Please submit all feedback by sending email to issues.
The downtime has ended
Posted: February 6, 2014
The networking changes have been completed and access to CHPC servers, file systems and VMs has been restored. Scheduling has been resumed on the clusters.
Please let us know if you see any issues
Major Change in Scope of Feb 6th Downtime -- Now 6am to 9am
Posted: February 3, 2014
Due to changes in the scheduling of the campus network changes and the electrical work in the Downtown Data Center, the scope of the downtime on Feb 6th has been changed. The CHPC networking changes and updates will be started at 6am. We estimate that all work will be completed by about 9am.
During most of this window, starting at 6am, the interactive nodes, any CHPC compute servers, the meteoXX and atmosXX boxes, as well as all VMs will be inaccessible. The batch schedulers on the clusters will be paused so that no new jobs can be started. Users will not have access to CHPC file systems (including CIFS/Samba mounted shares on desktops). Any existing connections to the VMs, interactive or compute nodes, and file systems may be lost during this window.
An announcement will be sent once the interactive nodes, VMs, and CHPC file systems are again accessible and the schedulers have been allowed to resume.
The campus networking outage to move INSCC to the new campus backbone has been tentatively rescheduled for Sunday, February 23rd.
The electrical work in the Downtown Data Center as well as any additional work on the clusters has been tentatively rescheduled for Tuesday, February 25th. During this February 25th downtime, CHPC will start the separation the smithp nodes from ember into a new cluster named ash.chpc.utah.edu in order to prepare for an upcoming expansion of this groups resources. This separation is necessary, due to limitations of the infiniband infrastructure, in preparation for a planned addition to Professor Smith's cluster. Once the new ash cluster is up, later in the week of the 25th, guest access will be allowed as before.
As we approach these dates, more information will be provided.
Spring CHPC Presentations Series
Posted: January 31, 2014
The Spring CHPC Presentation series will begin next week. Here is the schedule:
- Feb 6 Overview of CHPC
- Feb 11 Introductory Linux for HPC Part 1
- Feb 13 Introductory Linux for HPC Part 2
- Feb 18 Intro to Parallel Computing
- Feb 20 Protected Environment, AI and NLP Services **HSEB2908 at 2:00PM
- Feb 25 Using Gaussian09 and Gaussview at CHPC
- Feb 27 Intro to GPU Programming
- Mar 4 Python for Scientific Computing
All presentations are at 1:00PM in the INSCC Auditorium with the exception of the notation above. All are welcome to attend
XSEDE HPC Monthly Workshop - February 4, 2014 - Big Data
Posted: January 22, 2014
CHPC will host another XSEDE montly workshop on Tuesday, Feb. 4th focused on Big Data. This workshop will cover topics such as Hadoop and SPARQL.
The workshop will take place at INSCC Auditorium, Rm. 110, from 9am to 3pm MST.
To register, choose the University of Utah as a site on the XSEDE Portal Registration pages:
Please visit the workshop page for more information:
Downtime -- Change of Date -- Now Feb 6, 2014
Posted: January 22, 2014
Duration: Most of day
The date for the upcoming CHPC downtime has been changed to Thursday February 6, starting at 6:30am.
This downtime is necessary to make network routing changes on CHPC's part to complete our move out of SSB and Komas. This time will also be used by the campus to upgrade portions of the campus network that will impact INSCC and for electrical work at the DDC to allow for the addition of new gear. In addition, CHPC will use this downtime to further upgrade and update our network and to apply updates to the OS on the clusters.
If all goes as planned, the networking portion of the downtime will be complete by about 3:00PM.
Impact to users:
- Users will not have access to CHPC file systems (including CIFS/Samba mounted shares on desktops) until the networking changes have been completed
- Users in INSCC, Sutton and Browning will not have access to the internet at all for two periods of time - the first starting at about 8:00AM and the second at about 1:00PM
- Users will not have access to all CHPC compute servers (including kachina, swasey, database servers) and CHPC services such as websites and jira, until the networking changes have been completed
- Ember, Telluride, and the UCS (turretarch) nodes will be powered down; reservations are in place on Ember and Telluride to drain the queue by the start of the downtime
- The scheduler on kingspeak and apexarch will be paused until all networking tasks are complete, so that no new jobs can start
- Logins will be disabled on kingspeak, apexarch, as well as the meteo and atmos nodes until all networking changes are complete
CHPC Downtime - Feb 4, 2014 - starting at 6:30am
Posted: January 21, 2014
Duration: Most of day
The next CHPC downtime has been scheduled for Feb 4, 2014.
This downtime is necessary to make network routing changes on CHPC's part to complete our move out of SSB and Komas. This time will also be used by the campus to upgrade portions of the campus network that will impact INSCC and for electrical work at the DDC to allow for the addition of new gear.
In addition, CHPC will use this downtime to further upgrade and update our network and to apply updates to the OS on the clusters.
Due to the nature of the network outages, CHPC users can expect that they will not have access to CHPC file systems or servers for most of the day. More details will follow as we finalize the work schedule.
Updraft, sanddunearch, /scratch/general, /scratch/serial, and /scratch/uintah retirement - January 31, 2014 at 8am
Posted: January 17, 2014
As was announced last May, CHPC has been making plans to retire updraft, sanddunearch, /scratch/general, /scratch/uintah, and /scratch/serial. The date for the retirement of this equipment is January 31, 2014. Reservations are in place so the queues for the two clusters will be drained by 8am on January 31, 2014.
Any user that has data they need to keep on the retiring scratch file systems should move the data to either their home directories, group spaces, or to one of the remaining scratch file systems: /scratch/kingspeak/serial, /scratch/ibrix/chpc_gen, or /scratch/ibrix/icse.
The interactive nodes of updraft and sanddunearch as well as the scratch file systems will remain online until Feb 3rd at noon. The file systems will however be changed to read-only mode. This will provide a small window for users to migrate any data from the last jobs off of the scratch systems that will be retiring.
If you have any questions about this hardware retirement or need assistance with data migration, please let us know.
XSEDE HPC Workshop at CHPC - January 22, 2014 - OpenMP
Posted: January 7, 2014
Duration: All day, starting at 9am in INSCC Auditorium
This workshop is intended to give C and Fortran programmers a hands-on introduction to OpenMP programming. Attendees will leave with a working knowledge of how to write scalable codes using OpenMP.
Registration is required. Registration closes at 01/17/2014 17:00 EST. You can register at this link.
The tentative agenda is as follows - NOTE the times below are EST:
- 11:00 Welcome
- 11:15 Environment
- 11:30 It’s a Multi-core World
- 12:00 Intro to OpenMP
- 13:00 Lunch Break
- 14:00 Exercise 1
- 14:30 More OpenMP
- 15:15 Exercise 2
- 15:45 Advanced OpenMP Overview
- 16:00 Debugging & Profiling
- 16:30 OpenMP 4.0 Sneek Peek