HPC2N
High Performance Computing Center North
We just had a power outage this morning. All jobs running on both Akka and Abisko were interrupted. We are currently in the process of restoring the systems.
We hope to have all Akka compute nodes back in service by the end of today.
Abisko nodes may take longer. It is possible that we can get them back up today, but they may not be available until Thursday morning, due to the scheduled file system update tomorrow.
Tue, 2014-08-19 11:40 | Daniel Petersen
Quantum Espresso 5.1 has now been installed on both Abisko and Akka.
Thu, 2014-08-14 08:03 | Åke Sandgren
On Wednesday Aug 20th we will do the initial installation of our new cluster file system.
To do this we need to do some recabling on the interconnect of Abisko.
We will therefore drain the cluster of jobs so that it will be empty at 08:00
The system is estimated to be back in production on Thursday morning at the latest.
More information on the new file system will be annouced later.
Wed, 2014-08-13 10:39 | Åke Sandgren
Abisko will be upgraded to the latest version of Slurm on Wed. 2014-06-11.
No jobs will be running during the upgrade.
All jobs in the queue should be preserved and start as soon as the upgrade is completed.
Tue, 2014-06-03 17:13 | Magnus Jonsson
Akka will be upgraded to the latest version of Slurm on Tuesday 2014-06-03.
No jobs will be running during the upgrade.
All jobs in the queue should be preserved and start as soon as the upgrade is completed.
Tue, 2014-05-27 12:40 | Magnus Jonsson
The batch server for Akka will be move to another system on Mar 25th
To do this we have to make sure no jobs are running and have therefor created a reservation that starts 08:00 on the 25th.
We estimate that it will take about an hour to do the move.
Mon, 2014-03-24 10:37 | Åke Sandgren
We are currently experiencing some problem with the /pfs/nobackup file system on Abisko.
To get it back into working order we are forced to reboot some of the nodes, thereby killing the jobs on them (the jobs would fail anyway due to the problem)
We apologize for this and are working as fast as we can to get it back in working order.
*UPDATE 2014-03-21 18:33*
/pfs/nobackup is now (hopefully) back in working order on Abisko
Fri, 2014-03-21 19:16 | Åke Sandgren
Akka has now been (partially) upgraded to Ubuntu Precise.
The login node and some of the nodes are upgraded.
The rest will follow when running jobs are finished.
The environemt is now the same as it is on Abisko.
Documentation on the web has not been updated yet so please look at the examples for Abisko in the meantime and adapt them to fit the number of cores (8) per node on Akka.
Please notify support@hpc2n.umu.se in case of problems.
Starting on Monday Mar 17th Akka will be upgraded from the old CentOS distribution to the same Ubuntu 12.04 distribution that is running on Abisko.
Akka will also run slurm as the batch system.
This way both systems will look and feel the same (apart from differences in computing speed of course)
Submission of jobs have already been stopped.
Already submitted jobs will be allowed to finish.
Fri, 2014-03-14 16:47 | Åke Sandgren
As some of you may have noticed the /pfs/nobackup filesystem is chock full. As a side effect of that the batch systems are holding off on starting new jobs.
This is due to a lot of users keeping old files in the filesystem.
Please note that even if you are using a smaller amount than your quota limit you are obliged to remove old files and strive to keep your filesystem usage as small as possible.
