Because of an urgent security update the login node will be down for 1 hour.
Author Archives: lsz075
Hexagon: Updated software/libraries
Hexagon has updated software and libraries.
For full release notes from Cray, look at:
http://docs.cray.com/books/S-9401-1107//S-9401-1107.pdf
Note that some of the software listed there is not supported/will not be installed on hexagon.
Updates:
xt-mpt 5.3.0 -> 5.3.0
Petsc 3.1.05 -> 3.1.08
TPSL 1.0.01 -> 1.1.00
libfast 1.0.8 -> 1.0.9
ATP 1.2.0 -> 1.2.1
lgdb 1.3 -> 1.4
(xt-)gcc 4.5.3 -> 4.6.0
PGI 11.5.0 -> 11.6.0
xt-asyncpe 4.9 -> 5.00
For full release notes from Cray, look at:
http://docs.cray.com/books/S-9401-1107//S-9401-1107.pdf
Note that some of the software listed there is not supported/will not be installed on hexagon.
Updates:
xt-mpt 5.3.0 -> 5.3.0
Petsc 3.1.05 -> 3.1.08
TPSL 1.0.01 -> 1.1.00
libfast 1.0.8 -> 1.0.9
ATP 1.2.0 -> 1.2.1
lgdb 1.3 -> 1.4
(xt-)gcc 4.5.3 -> 4.6.0
PGI 11.5.0 -> 11.6.0
xt-asyncpe 4.9 -> 5.00
Hexagon: crash due to power-spike
Hexagon is down from 17:55 due to power-spike from thunderstorm. We are diagnosing and restarting the machine.
Update 20:25, machine is back up after disabling a failed module.
Update 20:25, machine is back up after disabling a failed module.
Hexagon: power down crash due to power-spike
Hexagon crashed/powered off due to a lightning strike and power-spike at 08:10.
Update 10:00: Cooling is also affected.
Update 12:47: Machine is up.
We used the time to include a future (planned) maintenance in this downtime.
Update 10:00: Cooling is also affected.
Update 12:47: Machine is up.
We used the time to include a future (planned) maintenance in this downtime.
Hexagon: Updated software/libraries
Hexagon has updated software/libraries. Due to static linking programs must be recompiled to incorporate fixes.
MPI
xt-mpt 5.2.3 -> 5.3.0
Libraries
TPSL 1.0.0 -> 1.0.01
Compilers and debuggers
PGI 11.4 -> 11.5
Intel 12.0.3.174 -> 12.0.4.191
Totalview 8.9.0 -> 8.9.1
lgdb 1.2 -> 1.3
ATP 1.1.3 -> 1.2.0
NOTES:
xt-mpt
Bug-fixes.
The following features were added to MPT 5.3.0 over MPT 5.2.3:
- Merged in the ANL MPICH2 1.3.1 release which includes numerous bug fixes
- Improved support for MPI thread safety. It is no longer necessary to link
in a separate library for thread multiple support. See the
MPICH_MAX_THREAD_SAFETY env variable in the intro_mpi man page for more info.
- Improve performance of MPI_Init when using static connections.
- Make several additional MPICH_GNI environment variables available to users
and document them in the intro_mpi man page.
TPSL
Bug fixed in TPSL 1.0.01:
772335 massive unnecessary diagnostic output from SuperLU_dist under
petsc/3.1.05
ATP and lgdb
Bugfixes.
MPI
xt-mpt 5.2.3 -> 5.3.0
Libraries
TPSL 1.0.0 -> 1.0.01
Compilers and debuggers
PGI 11.4 -> 11.5
Intel 12.0.3.174 -> 12.0.4.191
Totalview 8.9.0 -> 8.9.1
lgdb 1.2 -> 1.3
ATP 1.1.3 -> 1.2.0
NOTES:
xt-mpt
Bug-fixes.
The following features were added to MPT 5.3.0 over MPT 5.2.3:
- Merged in the ANL MPICH2 1.3.1 release which includes numerous bug fixes
- Improved support for MPI thread safety. It is no longer necessary to link
in a separate library for thread multiple support. See the
MPICH_MAX_THREAD_SAFETY env variable in the intro_mpi man page for more info.
- Improve performance of MPI_Init when using static connections.
- Make several additional MPICH_GNI environment variables available to users
and document them in the intro_mpi man page.
TPSL
Bug fixed in TPSL 1.0.01:
772335 massive unnecessary diagnostic output from SuperLU_dist under
petsc/3.1.05
ATP and lgdb
Bugfixes.
Fimm file system crashed
Fimm internal network crashed yesterday around 4:00 clock. and also we lost all file system, all running jobs are also crashed.
We are sorry for inconvenience.
We are sorry for inconvenience.
Hexagon: system crash
Hexagon crashed due to a power spike related to a thunderstorm which left cabinets in a fault state.
Update 15:12: System is up after 1 hour downtime.
Update 15:12: System is up after 1 hour downtime.
Fimm maintenance 30th May 2011
Dear Fimm Cluster Users :
We will have maintenance for fimm cluster from 08:00 ~ 16:00 on 30th May 2011(Monday).
Following will be performed :
* Add extra login node for fimm
* Reinstall cluster compute node
* Cable rearrangement
* Firmware update
During that time you will be able to access login node to perform basic operation, but you will not be able to submit any jobs or check queue status. Some of the file system will not be accessible or unstable during the maintenance.
Entire cluster is reserved for maintenance. All running jobs which will not be able to finish by the time the maintenance starts will be killed. user has to resubmit all killed jobs.
All submitted jobs which will not be able to finish by the time maintenance starts will be queued until the end of maintenance and will start running when maintenance is over.
If you have any further question please contact us at
hpc-support@hpc.uib.no
We are sorry for inconvenience.
Support team.
We will have maintenance for fimm cluster from 08:00 ~ 16:00 on 30th May 2011(Monday).
Following will be performed :
* Add extra login node for fimm
* Reinstall cluster compute node
* Cable rearrangement
* Firmware update
During that time you will be able to access login node to perform basic operation, but you will not be able to submit any jobs or check queue status. Some of the file system will not be accessible or unstable during the maintenance.
Entire cluster is reserved for maintenance. All running jobs which will not be able to finish by the time the maintenance starts will be killed. user has to resubmit all killed jobs.
All submitted jobs which will not be able to finish by the time maintenance starts will be queued until the end of maintenance and will start running when maintenance is over.
If you have any further question please contact us at
hpc-support@hpc.uib.no
We are sorry for inconvenience.
Support team.
Hexagon: Updated software/libraries
Hexagon has updates software/libraries.
MPT 5.2.3
Fixes:
- Correct handling of SHMEM locks at large core counts.
ATP 1.1.3
Fixes:
- When the Abnormal Termination Processing (ATP) signal handler recognizes that the application is running on a single core it does not initialize itself. ATP adds no value to single core applications. To take
advantage of this change, the application must be relinked.
- In previous versions of ATP, short running application that did not
fail could exit cleanly before that ATP was fully started which would result in atpFrontend processes hanging around indefinitely.
PGI 11.4.0
Features of PGI 11.4.0 are documented at: http://www.pgroup.com/doc/pgirn114.pdf
Fixes:
- 769940 pgcc-generated code for complex number operations inferior to Intel
GCC 4.5.3
Fixes:
- 768930 MANPATH for 'man gfortran' incorrect or 'old' version
- 769591 GNU OpenMP gfortran 4.5.2 internal compiler error for '!$omp task if(omp_get_num_threads() > 0)' [47886]
- 769876 Modulefile points to wrong location for man pages since v4.5.0
- 770495 PrgEnv-gnu does not set up correct man paths
- 770512 internal compiler error: in build_int_cst_wide, at tree.c:1178
- 771068 gcc module has the wrong manpath
- 771867 Incorrect MANPATH for gnu compiler man pages
Iobuf 2.0.2
IOBUF is an I/O buffering library that can reduce the I/O wait time for programs that read or write large files sequentially. IOBUF intercepts I/O system calls such as read and open and adds a layer of buffering, thus improving program performance by enabling asynchronous prefetching and caching of file data.
Fixes:
- BUG 771578 - iobuf module doesn't trap fwrite and fails when combined with posix
- BUG 772207 - Program not working when iobuf is being used
MPT 5.2.3
Fixes:
- Correct handling of SHMEM locks at large core counts.
ATP 1.1.3
Fixes:
- When the Abnormal Termination Processing (ATP) signal handler recognizes that the application is running on a single core it does not initialize itself. ATP adds no value to single core applications. To take
advantage of this change, the application must be relinked.
- In previous versions of ATP, short running application that did not
fail could exit cleanly before that ATP was fully started which would result in atpFrontend processes hanging around indefinitely.
PGI 11.4.0
Features of PGI 11.4.0 are documented at: http://www.pgroup.com/doc/pgirn114.pdf
Fixes:
- 769940 pgcc-generated code for complex number operations inferior to Intel
GCC 4.5.3
Fixes:
- 768930 MANPATH for 'man gfortran' incorrect or 'old' version
- 769591 GNU OpenMP gfortran 4.5.2 internal compiler error for '!$omp task if(omp_get_num_threads() > 0)' [47886]
- 769876 Modulefile points to wrong location for man pages since v4.5.0
- 770495 PrgEnv-gnu does not set up correct man paths
- 770512 internal compiler error: in build_int_cst_wide, at tree.c:1178
- 771068 gcc module has the wrong manpath
- 771867 Incorrect MANPATH for gnu compiler man pages
Iobuf 2.0.2
IOBUF is an I/O buffering library that can reduce the I/O wait time for programs that read or write large files sequentially. IOBUF intercepts I/O system calls such as read and open and adds a layer of buffering, thus improving program performance by enabling asynchronous prefetching and caching of file data.
Fixes:
- BUG 771578 - iobuf module doesn't trap fwrite and fails when combined with posix
- BUG 772207 - Program not working when iobuf is being used
Hexagon: Updated software/libraries
Hexagon has updates software/libraries.
MPI
xt-mpt 5.2.1 -> 5.2.2
Compilers, wrappers and debuggers
PGI 11.2.0 -> 11.3.0
Intel 12.0.2.137 -> 12.0.3.174
Chapel 1.2.1 -> 1.3.0
Totalview 8.8.0 -> 8.9.0
Java 1.6.0-22 -> 1.6.0-24
Libraries
xt-libsci 10.5.01 -> 10.5.02
NOTES:
xt-mpt
Bugs fixed in this release:
767853 Add error message for hitting request limit on Seastar instead
of segfault
xt-libsci
LibSci 10.5.02 includes bugfixes.
Bug 769615 - scalapack routine pdsyev aborts when global matrix > sqrt (2^31)
Intel
The following bugs are fixed in the Intel 12.0.3.174 release.
767152 ifort OpenMP atomic subtraction produces incorrect answers [611742]
Java
Security update
MPI
xt-mpt 5.2.1 -> 5.2.2
Compilers, wrappers and debuggers
PGI 11.2.0 -> 11.3.0
Intel 12.0.2.137 -> 12.0.3.174
Chapel 1.2.1 -> 1.3.0
Totalview 8.8.0 -> 8.9.0
Java 1.6.0-22 -> 1.6.0-24
Libraries
xt-libsci 10.5.01 -> 10.5.02
NOTES:
xt-mpt
Bugs fixed in this release:
767853 Add error message for hitting request limit on Seastar instead
of segfault
xt-libsci
LibSci 10.5.02 includes bugfixes.
Bug 769615 - scalapack routine pdsyev aborts when global matrix > sqrt (2^31)
Intel
The following bugs are fixed in the Intel 12.0.3.174 release.
767152 ifort OpenMP atomic subtraction produces incorrect answers [611742]
Java
Security update