vassilios.mewes at uv.es
Wed Feb 5 11:16:44 CST 2014
On Wed, Feb 5, 2014 at 6:02 PM, Ian Hinder <ian.hinder at aei.mpg.de> wrote:
> A "simulation", as far as simfactory is concerned, is a directory, which
> might be in /scratch/username/simulations. This usually corresponds to a
> single parameter file, and is logically "one cactus run". However, because
> jobs on clusters usually are limited to 24 hours, you need multiple jobs to
> form a longer simulation. SimFactory calls these different segments of the
> simulation "restarts", as the second one is restarted from the first, etc.
> The restarts correspond to directories output-NNNN in the simulation
> directory. Each new Cactus job in the simulation is run inside this
> output-NNNN directory. The parameter file usually specifies the name of
> the directory to put data into. Some users use '$parfile' which names the
> output directory after the parameter file.
> I usually name the simulation after the parameter file, and output the
> data into a directory also named after the parameter file, so my data would
> go in
> for a simulation called bbh, and an output file called phi.x.asc.
> Does this make things clearer?
yes, thanks for the explanation!
So in that case, you would set
IO::out_dir = bbh
IO::checkpoint_dir = bbh
IO::recover_dir = bbh ??
And another question about the restarts of the entire simulation: Do I need
to use the TerminationTrigger thorn for that and set the walltime I have on
the cluster (minus some time to do the checkpointing before the termination)
sorry for all the stupid questions, but this is the first time I will
actually have to use the simfactory!
> best wishes,
> On Wed, Feb 5, 2014 at 5:27 PM, Erik Schnetter <schnetter at cct.lsu.edu>wrote:
>> On Feb 5, 2014, at 11:19 , Roland Haas <rhaas at tapir.caltech.edu> wrote:
>> > Signed PGP part
>> > Hello Vassili,
>> > > sorry to keep bothering you...
>> > No problem. Sorry for the delayed reply, I had not seen your second
>> > question in the email.
>> > > I tried around getting the output folder to be in the
>> > > simulation_name/SIMFACTORY/data folder, but couldn't figure it
>> > > out..its not terribly important, but would be nice to write the
>> > > data to that folder..
>> > >
>> > > any ideas?
>> > Sorry, I have no idea how to change the output directory. I suspect
>> > that the output-0000 names are hard-coded into simfactory. In
>> > particular I do not think it is possible (nor advisable) to accumulate
>> > output from different segments in the same directory.
>> I recommend against outputting several segments into the same directory.
>> This seems convenient at first, but can completely destroy existing output
>> if you e.g. run out of disk space. In particular, if an HDF5 file is opened
>> for appending, it becomes unreadable until it has been properly closed, and
>> running out of disk space prevents this.
>> If you want to do so anyway, then you would use an output directory
>> "../data" or so. The path ".." gets you out of the output-NNNN directory.
>> I also would not use the SIMFACTORY directory; this directory contains
>> internal data for Simfactory, and is used to hide these data. I would use
>> e.g. simulation_name/data instead.
>> Erik Schnetter <schnetter at cct.lsu.edu>
>> My email is as private as my paper mail. I therefore support encrypting
>> and signing email messages. Get my PGP key from http://pgp.mit.edu/.
> Users mailing list
> Users at einsteintoolkit.org
> Ian Hinder
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Users