[MITgcm-devel] [MITgcm-support] reducing the size of STDOUT files etc

[Jean-Michel Campin]

Hi Dimitris,

I don't know if we should advertise too much this CPP option:
the effect is to discard any message (error/warning ...) from
any processor different from 0 (it will not transfert the error
message to STDERR.0000 but just drop it).
Now, many error messages are detected by all procs, but it's far to
be a general rule.

And before turning on this option, it seems to me that everything else
need to be checked (such as, in this case, monitorFreq) including 
how much wall clock time is really lost when all procs opening & write 
to STDOUT+STDERR.
The fact is, with the 3 suggestions you listed here:
http://mitgcm.org/pipermail/mitgcm-support/2009-April/006013.html
(+ also not to set debugMode=TRUE in eedata)
there is not much written in STDOUT after the initialisation is over.

I propose to change (again) the description of SINGLE_DISK_IO
in CPP_EEOPTIONS.h, by adding:
< C WARNING: to use only when absolutely confident that the setup is working
< C     since any message (error/warning/print) from any proc <> 0 will be lost.
but may be we should also add a warning in STDERR.0000 if this is turned on.

Cheers,
Jean-Michel

On Mon, May 12, 2014 at 02:32:42PM +0000, Menemenlis, Dimitris (329D) wrote:
> There is also a CPP_EEOPTIONS.h flag that can be used to write
> out a single STDOUT and STDERR file:
> 
> C--   Flag defined for eeboot_minimal.F, eeset_parms.F,
> C     and open_copy_data_file.F to write STDOUT, STDERR,
> C     and scratch files from process 0 only.
> #define SINGLE_DISK_IO
> 
> On May 12, 2014, at 4:58 AM, Jonny Williams <Jonny.Williams at bristol.ac.uk<mailto:Jonny.Williams at bristol.ac.uk>> wrote:
> 
> Hi there
> 
> I am trying to reduce the size of the output from my model because I am running on a very large number of processors and all the STDOUT* files (for example) are taking up too much space.
> 
> I've followed the instructions on this MITgcm support page...
> 
> http://mitgcm.org/pipermail/mitgcm-support/2009-April/006013.html
> 
> ... but I am still getting a STDOUT file for every processor (>1000) which is killing my disk space allowance! Is there something that I've missed? I thought that running with useSingleCpuIO=.TRUE. should mean that I only get 1 STDOUT file? I've attached my data file in case anyone has any ideas as to what is going wrong?!
> 
> Many thanks for your help.
> 
> Jonny
> 

> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support