[Mesa-users] Disabling eos, kap, etc. caching
W.H.Ball at bham.ac.uk
Thu Feb 7 11:56:15 EST 2019
Sadly, I'm not the administrator and I don't know of any scratch hard
drive space that I could use. Even if they could all read from another
user's cache, that one user would still have this issue. And I'm also not
sure that's even possible because of how the different filesystems with
user spaces are being mounted.
There are various other sysadmin ways of solving the problem but I was
hoping for a way to do it within MESA...
Postdoc, School of Physics and Astronomy
University of Birmingham, Edgbaston, Birmingham B15 2TT
W.H.Ball at bham.ac.uk
+44 (0)121 414 4552
On Thu, 7 Feb 2019, aaron.dotter at gmail.com wrote:
> Hi Warrick,
> See the section on cache directories in star_job.defaults. You might set one up for them all to use. Have you tried that?
> On Thu, Feb 7, 2019 at 11:43 AM Warrick Ball via Mesa-users <mesa-users at lists.mesastar.org> wrote:
> Hi everyone,
> I have a group of undergraduate students who are running MESA on a shared
> server with just 1 GB of home directory space per student. MESA itself
> has been installed in a shared (and root-only accessible) area, so it
> doesn't count against their quotas. However, after running a roughly
> solar-mass track to near the helium flash, many of them are approaching or
> filling their 1 GB quotas. It turns out this because of all the cache
> files being written to `.mesa_temp_cache` under their MESA work folder.
> Is it possible to stop or limit the use of the cache?
> As a temporary fix, I've advised them to delete cached files manually as
> necessary. I had a look for a control but it seems that the option of
> whether or not to use the caches is hardcoded. Grepping for "cache" and
> following some function calls, I found this at line 401 (MESA r10398) in
> logical, parameter :: use_cache = .true.
> `use_cache` is then passed to `init` routines for `eos`, `kap`, `atm` and
> `ionization`. Besides changing this hardcoded value to `.false.`, I can't
> see an option. Also, I can't see that it's easy to add an option to the
> inlists because of the sequence of the initialisation calls. This line is
> ultimately called through `star/public/star_lib:starlib_init`, which is
> part of the `before_evolve_loop` subroutine in
> `star/job/run_star_support.f90`. It's called *before* `star_setup`, which
> seems to be what reads the namelist.
> Alternatively, is there a way to limit the amount of space the caches use?
> Or tell MESA to clear the caches after some number of steps? Perhaps a
> control like `clear_cache_interval`?
> Warrick Ball
> Postdoc, School of Physics and Astronomy
> University of Birmingham, Edgbaston, Birmingham B15 2TT
> W.H.Ball at bham.ac.uk
> +44 (0)121 414 4552
> mesa-users at lists.mesastar.org
More information about the Mesa-users