[Japanese]
We manage a mailing list for announce concerning supercomputer, ml-super . A new user automatically participates. If you find you are missed, please contact us. You can freely escape from the list, but you will do not get announcement.
You can participate even if you are not an user. Not only for getting announcement, you can use it for communications with other users.
To attend and escape, please mail to consult@rcnp.osaka-u.ac.jp .
The following is a notice for the schedule of stopping systems due to stops of electricity, maintenances, and so on.
Since the supercomputer system is cooperatively managed by RCNP and Cybermedia Center, please be aware of the schedule of both the centers.
The following occasions appear according to the stopped systems.
In the cases of scheduled stops, job is restarted when the system comes back from the check-point at the down. You do not need to resubmit jobs. Only keep to untouch the files during I/O.
There are craches of single CPU and whole a node. In the case of crash of single CPU, only the job on that CPU abnormally finishes. If the whole node crashes, all the jobs on that node finishes abnormally. In the case of a parallel job, if even one of the distributed processes crashes, whole the job finishes abnormally. In any case, jobs before execution receive no effect.
Abnormally finished jobs are re-run in the default setting. If you do not prefer re-run, please set the option of qsub -nr (no re-run).
The list of abnormally terminated jobs will be announced within a day. (We are sorry that we do not announce to each user.)