blog

Total number of slots consumed in long_excl.q (exclusive nodes) will be restricted to 320 slots on June 4, 2013

Tue, 28 May 2013 16:14:01 -0700

The long queue has had a per-user limit of 320 slots (or 40 8-core nodes) for several months. The long_excl.q has not had a similar limit. On June 4, a per-user 320 slot limit will be introduced on the exclusive nodes (or 20 16-core nodes). This will ensure that a single user cannot consume too many resources simultaneously in the long queues. For large calculations requiring a great deal of resources, the normal queues should be targeted (jobs < 12 hours), as there are no per-user slots limits on those queues.

New Hardware to be Integrated by Jan 7, 2013

Thu, 27 Dec 2012 14:57:54 -0800

The new hardware that has been tested through phoebe will be available on Genepool no later than January 7, 2013. The following compute power will be added to Genepool: 220 nodes with 16 physical cores and 120G of memory; 5 nodes with 32 physical cores and 500GB of memory; 2 nodes with 32 physical cores and 1000GB of memory.

Genepool Quarterly Maintenance

Wed, 07 Nov 2012 10:45:39 -0800

The Genepool cluster will be offline for maintenance next Tuesday, November 13th from 7:00 am to 7:00 pm (PST). During this time staff will upgrade the Genepool compute nodes to Debian 6. Also, the firmware for /projectb and /house will be updated. /projectb will be unavailable from 7:00 am to 7:00 pm (PST) and /house will be unavailable from 8:00 am to 1:00 pm (PST). Any webservices or workflows that depend on these file systems will also be unavailable. Please plan your work accordingly.

Rolling Debian 6 Updates for Interactive Machines and Web Servers 11/2 - 11/9

Wed, 07 Nov 2012 08:05:58 -0800

The JGI computational resources are being updated to Debian 6 this week and next. Because of the staff-intensive work required to update the login nodes, interactive nodes, and the web servers, these machines will be updated between 11/02 and 11/09. The genepool cluster nodes will be updated during the scheduled quarterly maintenance on 11/13 between 7:00AM and 7:00PM.

The work on the interactive nodes will begin each day at 9:00AM, and should conclude by 5:00PM.

Debian6 Update Schedule 11/02/2012 - 11/09/2012

Friday, 2012-11-02

gpint01 - Previously down for hardware maintenance *complete
sdm-pacbio / gpint23 *complete
rqc-dev / gpint24 *complete
gpweb07 / img-edge4 *complete
gpweb13 -> gpint15 w/ Debian6 Update *complete

Monday, 2012-11-05

gpint07 / ranger - Plant gpint *complete
gpint14 - general purpose *complete
gpint19 - general purpose *complete
gpint20 - general purpose *complete
gpweb04 / img-edge1 *complete
gpweb11 / frylock *complete

Tuesday, 2012-11-06

genepool01 - login node *complete
gpint03 / bcg2 - Vista gpint *complete
gpint08 / boiler - Plant gpint *complete
gpint16 - general purpose *complete
gpint21 - general purpose *complete
gpweb09 / galaxy-edge *complete
gpweb10 / galaxy-dev *complete
gpweb12 / gp-next *complete

Wednesday, 2012-11-07

genepool02 - login node *complete
gpint05 / img-worker - IMG gpint *complete
gpint09 / sedona - Plant gpint *complete
gpint17 - general purpose *complete
gpint22 - general purpose *complete
uhura *complete

Thursday, 2012-11-08

genepool03 - login node
gpint04 / merced - IMG gpint
gpint10 / willow - Plant gpint
gpint12 / bat - R&D gpint
gpint18 - general purpose
gpweb05 / img-edge2
gpweb06 / img-edge3

Friday, 2012-11-09

genepool04 - login node
gpint06 / zeus - IMG gpint
gpint11 / actinium - Plant gpint
gpint13 / quarter - Fungal gpint
gpint02 / bcg1 - Vista gpint
gpweb01 / vista - genome.lbl.gov (among others)
gpweb02 - hazelton

Debian 6 testing on Phoebe - complete by Nov 1st.

Mon, 01 Oct 2012 16:51:43 -0700

NERSC is planning to install Debian 6 as the OS on Genepool at the November quarterly maintenance. To prepare for this, all users need run and test pipelines on Phoebe. A testing schedule has been put on a google doc here:

https://docs.google.com/a/lbl.gov/spreadsheet/ccc?key=0AlQ6Zwxa5cirdC1hM3NHMk5IVkMycnVDdkN0OF94LXc#gid=0

Rolling gpint upgrades 10/2-10/10

Mon, 01 Oct 2012 16:46:32 -0700

NERSC needs to update-and-correct the system software installed on the gpint machines. To achieve this, we need to schedule downtime for each of the gpints. The downtime will last from 9:00AM - 5:00PM on the scheduled date. One purpose of this work is to adjust the installed Debian 5 software packages to bring all genepool and genepool-attached systems to a single common OS image. This will help to ensure that installed software on genepool will work correctly on the gpints - there are several cases in open tickets where software works on the genepool login nodes but fails on the gpints; the will also facilitate easier maintenance of the gpints. Another purpose is to perform some hardware maintenance of the gpints where required.

*Public Access gpint Machines*

gpint16: October 2, 2012
gpint17: October 3, 2012
gpint18: October 4, 2012
gpint19: October 2, 2012
gpint20: October 8, 2012
gpint21: October 9, 2012
gpint22: October 10, 2012

*Group gpint Machines*
gpint01: October 2, 2012
gpint02: October 3, 2012
gpint03: October 4, 2012
gpint04: October 5, 2012
gpint05: October 8, 2012
gpint06: October 9, 2012
gpint07: October 10, 2012
gpint08: October 2, 2012
gpint09: October 3, 2012
gpint10: October 4, 2012
gpint11: October 5, 2012
gpint12: October 8, 2012
gpint13: October 9, 2012
gpint14: October 10, 2012

JGI Computational Resources Quarterly Maintenance Tuesday August 14, 7AM - 6PM

Fri, 03 Aug 2012 12:09:11 -0700

Dear JGI Community,

It is now getting close to the quarterly maintenance of the servers, shared filesystems, and HPC clusters operated by NERSC, which is scheduled for all day on Tuesday August 14, 2012. This maintenance will include a shutdown of the /house and /ifs Isilon filesytems which means that some of your desktop systems may not function as they usually do. If you have a Linux workstation, your home directory is most likely hosted on /house and you may not be able to successfully login. Windows and Mac users will be less effected by this maintenance - though will still be unable to access files and folders hosted on /house. There will be a period of network maintenance which will disable network connectivity between NERSC and JGI, during that time some JGI websites may be unavailable. Also, some of the JGI websites may be affected by filesystem changes throughout the day.

The High Performance Computing clusters, genepool, phoebe and kronos, will be unavailable throughout the day on the 14th. Jobs that are submitted and queued but cannot run before the maintenance will be retained and scheduled to run afterwards. To determine the maximum time limit for a job to run before the maintenance - type "countdown" on your genepool or gpint prompt.

You are welcome to email NERSC at consult@nersc.gov if you have any questions about the maintenance, or its affects on your work.

Sincerely,

Doug Jacobsen

NERSC Bioinformatics Computing Consultant

Planned Maintenance Schedule

Genepool Cluster (8:00am - 6:00pm August 14, 2012):
OS and UGE upgrades, also offline through /house maintenance

Isilon (8:00am - 4:00pm August 14, 2012):
Access to the Isilon (/house) will be interrupted for approximately 8 hours while the operating system is upgraded.

/home/public_web relocation (9:00am - 9:30am, 4:00pm - 6:00pm August 14, 2012):
Several JGI websites will be temporarily impacted by the 9-9:30am network maintenance, and down for docroot relocation to House.

Server Moves (7:00am - 6:00pm August 14, 2012):
The following systems will be physically relocated from JGI to OSF:
sdm-dev, zome-private, rqc-dev, portal, sdm-prod, rqc-sdm-edge1, gp-next, gp-edge3, rqc-sdm-edge2, rqc-prod, gp-edge4, jel, anu

Network Maintenance (9:00am - 9:30am August 14, 2012)
All JGI <-> NERSC traffic down
Public traffic to gp*.nersc.gov down

Misc Server maintenance (8:00am - 6:00pm August 14, 2012)
dover, gpint05, gpint21, iceland

Python Boot Camp - Aug. 20-22nd at UCB

Mon, 30 Jul 2012 16:20:22 -0700

JGI Users,

UC Berkeley is offering a free 3 day Python boot-camp on August 20-22nd. The boot camp is open to the LBL community. To register please see the following page:

http://register.pythonbootcamp.info/home

ParLab Boot Camp 2012 Aug 15-17th

Tue, 03 Jul 2012 10:41:30 -0700

JGI Users are highly encouraged to attend the UC Berkeley's ParLab Boot Camp on parallel and mulit-core programming. The free workshop is open to all LBL employees and is hosted by UCB, the ParLab and NERSC. It is a 3 day hands-on boot camp with top speakers from around the country.

Registration will be open in Mid-July. See the ParLab webpage for more information.

http://parlab.eecs.berkeley.edu/2012bootcamp

/projectb now mounted on Carver

Thu, 07 Jun 2012 22:59:21 -0700

The /projectb file system has been mounted on Carver.

Old shares set to zero, default projects set

Thu, 07 Jun 2012 22:58:30 -0700

Each user now has a default share project. If you'd like your default changed, please send email to consult@nersc.gov or file a ticket with help.nersc.gov.

Old project shares have been set to zero. Please make sure to use the new shares.

New Shares/Projects Available on Genepool

Thu, 24 May 2012 12:35:58 -0700

Users are now able to request the new projects specified here:Queues and Policies , when running on Genepool.

How to get help from NERSC: User Training May 9th 1:30pm

Thu, 26 Apr 2012 13:22:40 -0700

On May 9th, we will hold a training session for users on how to get help from NERSC. Front line JGI user support will be transitioned to the NERSC consulting team within the next month. At this training we will show you how to contact the NERSC consultants and how to track the progress of your problem reports. We will also explain our service model and service level commitments and goals.

When: May 9th 1:30-2:30pm
Where: Room 149 A/B
What to bring: Your laptop

NetApp file systems unmounted May 8th

Thu, 26 Apr 2012 13:20:58 -0700

As a reminder, the NetApp file systems are being retired. They are currently set to read-only and on May 8th the mount point will be changed. Users should make sure that all important data in the NetApp file systems has been archived to HPSS or migrated to the /house or /projectb file systems. Users will still have access to the NetApp file systems from the data transfer nodes until we are confident that all data has been backed up.

Why is my job NOT running? User Training April 18th

Wed, 18 Apr 2012 11:27:21 -0700

This training will be very hands-on, so please bring a laptop. We will be solving user questions as a group so everyone can see how to diagnose some of the common problems. You will be able to get one on one assistance if necessary. We will also cover and assist with data migration from /home to /house.

Where: 100-149 A/B
When: Wednesday, April 18th from 1:30-3:30 (or as long as people have questions)

Training: "Why isn't my job running?"
Agenda:
(1:30-2:30)
- Overview of queues and complexes, as well as the Genepool configuration (15 min)
- Basic submission options for qsub - "Where will my job end up?" (10 min)
- Open up the floor to let anyone ask about jobs that aren't running. We'll try to solve your problems and answer questions in real-time in front of the group. This will highlight the use of the qstat and qconf commands. (30-40 min, possibly longer)

(2:30-3:30)
- Overview of environment variables and modules (20 min)
- More advanced job submission information - array jobs (10 min)
- Open up the floor to more questions about workflows and adapting submission scripts to include job array syntax.

- Data transfer - Netapps are going read only, are there people here who need help transferring their data

(3:30 -?)
- General help session (until the room is empty)

Crius Decommission

Tue, 10 Apr 2012 18:10:02 -0700

The Crius system is in the process of being decommissioned. All groups, besides the fungal/annotation group are no longer able to submit jobs to Crius. The fungal group is finishing up testing on Genepool. 40 Crius nodes are being moved to Genepool now and the final 40 nodes will be moved when fungal's testing is completed.

NetApps file system set to read-only April 17th

Fri, 06 Apr 2012 14:50:10 -0700

On April 17th we plan to set the NetApps file system to read-only. This is in preparation for retiring the NetApps file system at the end of May. Please make plans to archive this data to NERSC's HPSS tape system or move them to another file system like house or projectb.

After April 17th you will be able read data from the NetApps, but not write to them. Information on moving data to NERSC's tape archive is located here.
https://www.nersc.gov/users/computational-systems/genepool-phoebe/transfer-and-archive-data/

If you need help or have concerns about this change please let us know as soon as possible. The NetApps file system is already off maintenance protection and we are pushing to get all user data off of them because at this point, we would be unable to recover any data from the NetApps file system in case of failure.

Create a virtual desktop with NX - Tutorial Wed April 11th 1:30pm

Thu, 05 Apr 2012 13:29:59 -0700

Tutorial Wedneday April 11th at 1:30pm in rm 100-149A/B. Bring your laptop.

NX is a computer program that handles remote X Window System connections and it provides two benefits for NERSC users:

NX can greatly improve the performance of X Windows to the point that it can be usable over a slow link such as a dial-up modem.
NX provides sessions that allow a user to disconnect from the session and reconnect to it at a later time while keeping the state of all running applications inside the session. This gives users a virtual desktop that's running at NERSC.

high.c resource available for debugging, testing and high priority jobs

Wed, 04 Apr 2012 17:06:33 -0700

The high.c resource is available for users who need fast turn around for debugging, testing or high priority jobs. There is a limit of 12 hours and 2 slots for the high.c resource.

Queue Resources Deprecated

Wed, 04 Apr 2012 09:57:21 -0700

long.c, crius_bg.c, crius_high.c, crius_normal.c, galaxy_bg.c, galaxy_high.c, galaxy_normal.c, medium.c, short.c

The following resources will be deprecated April 4th.