Difference between revisions of "UKCA Chemistry and Aerosol vn10.9 Tutorial 1"

From UKCA
Line 63: Line 63:
 
==Output Directory Structure==
 
==Output Directory Structure==
   
[[Image:Cylc-run_structure.png|thumb|400px|left|Simplified cylc-run directory structure. '''[SUITE-ID]''' will be e.g. '''''u-as101''''', and '''[CYCLE]''' is the cycle-point, e.g. '''''1''''' or '''''19880901T0000Z'''''.]]
+
[[Image:Cylc-run_structure.png|Simplified cylc-run directory structure. '''[SUITE-ID]''' will be e.g. '''''u-as101''''', and '''[CYCLE]''' is the cycle-point, e.g. '''''1''''' or '''''19880901T0000Z'''''.]]
   
 
The output directory structure of rose suites are rather complex. The schematic on the right gives a broad overview of the general structure. Everything can be found within the '''<code>[SUITE-ID]</code>''' (e.g. ''u-as101'' etc.) directory, which can be found within your '''<code>$HOME/cylc-run</code>''' directory. On the VM, everything in held in this single place, but things are more complicated when using PUMA & ARCHER, as some files are held on the ARCHER ''<code>/home</code>'' filesystem, some on the ''<code>/work</code>'' filesystem, and some files are synced back to PUMA as well.
 
The output directory structure of rose suites are rather complex. The schematic on the right gives a broad overview of the general structure. Everything can be found within the '''<code>[SUITE-ID]</code>''' (e.g. ''u-as101'' etc.) directory, which can be found within your '''<code>$HOME/cylc-run</code>''' directory. On the VM, everything in held in this single place, but things are more complicated when using PUMA & ARCHER, as some files are held on the ARCHER ''<code>/home</code>'' filesystem, some on the ''<code>/work</code>'' filesystem, and some files are synced back to PUMA as well.

Revision as of 08:07, 19 December 2017

UKCA Chemistry and Aerosol Tutorials at vn10.9

Before starting these tutorials you should first make sure that you have completed the setup instructions. If you are using Linux or macOS you will only need to complete section 2.2 onwards.

Copying and Running an Existing Rose Suite

Machine UM Version/Configuration Suite ID
ARCHER vn10.9 N48L38 Intel Compiler u-as101
vm vn10.9 N48L38 GNU Compiler u-as159

If you are using PUMA & ARCHER, you will need to login to PUMA, e.g.

ssh -Y username@puma.nerc.ac.uk

If you are using the Met Office Virtual Machine, you will need to login to the VM, e.g.

vagrant ssh

You should then be asked for your SRS password.

Then launch the UM graphical user interface by:

rosie go
Figure 1: GCylc running.

This should then load up a blank interface. Go to Edit Data source and select u. The go to the search panel and search for the correct base suite. This will then show all the suites that have this suite-id in it's history or title. You should just select the correctly named suite and not any of the others.

Right-click on the suite and click copy suite. A new box will open asking for the project - it is fine to press Forward here. On the next panel, it's fine to just press OK again as well. You should avoid the checkout suite option, as this will give you a copy of someone else's suite, and it might be possible for you to make changes to it that could affect other users.

The suite will now copy and checkout to your /home/$USER/roses directory, and will also appear when running rosie go.

You should now right-click edit (or double-click) on this new suite, and click run, which is symbolised by a play symbol (i.e., a large grey arrow-head pointing to the right).

If you are using PUMA/ARCHER, this suite has been configured to manually compile, so you should follow the instructions on how to do that, to be sure that you can compile your suite as quickly as possible. It will take about 7 minutes to compile the code (fcm_make and fcm_make2 tasks), followed by about 3 minutes to run the reconfiguration step (recon), and then 1 minute to run the UM itself (atmos).

If you are using the VM, times will vary depending on the specifications of the host. It could take about 10 minutes to compile the code (fcm_make) for the first time (and about 1-2 minutes when recompiling), followed by about 1 minute to run the reconfiguration step (recon), and then about 12 minutes to run the UM itself (atmos).

When the suite has finished successfully it will then become blank with the message stopped with 'succeeded' in the bottom-left corner.

Version Control

Rose suites are all held under version control, using fcm. When making changes to a suite, you will need to save it before you can run the suite. Once you are happy with the settings, you can also commit these changes back to the repository - to do this change directory to the

/home/$USER/roses/[SUITE-ID]

and then type

fcm commit

a text editor will then open, and you should type a short message describing what the changes you have made do. You should then close the editor and type y in the terminal. It is recommended that you commit frequently (even on configurations that aren't working) as this protects you against mistakes and accidental deletions etc.

These suites can be viewed on the SRS here: https://code.metoffice.gov.uk/trac/roses-u (password required)

It is recommended that you commit your suites regularly.

Output Directory Structure

Simplified cylc-run directory structure. [SUITE-ID] will be e.g. u-as101, and [CYCLE] is the cycle-point, e.g. 1 or 19880901T0000Z.

The output directory structure of rose suites are rather complex. The schematic on the right gives a broad overview of the general structure. Everything can be found within the [SUITE-ID] (e.g. u-as101 etc.) directory, which can be found within your $HOME/cylc-run directory. On the VM, everything in held in this single place, but things are more complicated when using PUMA & ARCHER, as some files are held on the ARCHER /home filesystem, some on the /work filesystem, and some files are synced back to PUMA as well.

Within this directory there several directories, including:

  • log: a symbolic link to a directory called something like log.20171214T141332Z, which contains all the job.out (containing output from UM WRITE statements from the umPrint subroutine) and job.err files etc., as well as the script used to run the job.
  • work: this contains the directories used when the job actually runs. Real-time output will be sent to files here (held in a pe_output directory. Sometimes model output will also be here. On ARCHER this directory exists on the /work filesystem.
  • share: This isn't used by the UKCA Training Suite, but for climate jobs model output will usually be held in this directory. On ARCHER this directory exists on the /work filesystem.

Within these directories there will be many sub-directories. Some of these will be named from the cycle-point (labelled [CYCLE] in the graphic. For the UKCA Training Suite this will be 1, but under normal climate operation this will be a date string, e.g. 19880901T0000Z.

Files General Path Example
Most recent job.out files ARCHER:/home/n02/n02/$USER/cylc-run/[SUITE-ID]/log/job/1/[JOB NAME]/NN
vm: /home/vagrant/cylc-run/[SUITE-ID]/log/job/1/[JOB NAME]/NN
/home/n02/n02/luke/cylc-run/u-as292/log/job/1/atmos/NN
/home/vagrant/cylc-run/u-as297/log/job/1/atmos/NN
(These can also be viewed through the Gcylc GUI right-click menu from each step.)


  • You can find the latest & full job.out files in
ARCHER:/home/n02/n02/$USER/cylc-run/[SUITE-ID]/log/job/1/[JOB NAME]/NN
e.g. /home/n02/n02/luke/cylc-run/u-as292/log/job/1/atmos/NN
These are also mirrored to PUMA, in your /home/$USER/cylc-run/[SUITE-ID] directory.
vm: /home/vagrant/cylc-run/[SUITE-ID]/log/job/1/[JOB NAME]/NN
e.g. /home/vagrant/cylc-run/u-as297/log/job/1/atmos/NN
This file contains what the UM writes out using the umPrint subroutine. These can also be viewed through the Gcylc GUI right-click menu from each step.
  • Output from the individual processors (which are updated as the job is running can be found in
ARCHER: /work/n02/n02/$USER/cylc-run/[SUITE-ID]/work/1/[JOB NAME]/pe_output
e.g. /work/n02/n02/luke/cylc-run/u-as292/work/1/atmos/pe_output
vm: /home/vagrant/cylc-run/[SUITE-ID]/work/1/[JOB NAME]/pe_output
e.g. /home/vagrant/cylc-run/u-as297/work/1/atmos/pe_output
A handy command to check the progress of the job is
ARCHER: tail -1000f atmos.fort6.pe00 | grep Atm_Step
vm: tail -1000f atmos.fort6.pe0 | grep Atm_Step
  • Output files in 64-bit fieldsfile format can be found in:
ARCHER: /work/n02/n02/$USER/cylc-run/[SUITE-ID]/work/1/atmos
e.g. /work/n02/n02/luke/cylc-run/u-as292/work/1/atmos
Note: other suites may put the output into a different location, e.g. /work/n02/n02/$USER/cylc-run/[SUITE-ID]/share/data/History_Data, and there are also post-processing settings that can be used to copy data to the RDF or JASMIN as the suite runs.
vm: /home/vagrant/cylc-run/[SUITE-ID]/work/1/atmos
e.g. /home/vagrant/cylc-run/u-as297/work/1/atmos
Files can have various naming conventions. For these suites, they will be atmosa.pa19810901_00, but more generally they are likely to be of the form
[SHORT SUITE-ID]a.p[abcdefghijkmsyx]YYYYMMDD
e.g. ag308a.pk19880901
[SHORT SUITE-ID]a.p[abcdefghijkmsyx]YYYYmon
e.g. ak468a.pe1989oct

Whilst this suite is running, take a look at Tutorial 2: Exploring Rose.

Viewing Output

Example output from the UKCA training suite.

For more detailed plotting, tools such as the Iris and cf-python libraries can be used to view UM file formats directly. However, for quick viewing, Xconv is a very useful tool.

To view these files, do

xconv atmosa.pa19810901_00

As well as viewing files, you can use Xconv to convert these files to netCDF, by filling in the Output file name: box (e.g. foo.nc), and then clicking convert. If no path is defined, this will save the file in the same directory that you opened Xconv from.

Example output from the UKCA training suite can be found at /work/n02/n02/ukca/Tutorial/vn10.9/sample_output/Base/atmosa.pa19810901_00 on ARCHER.


Checklist

List suites using rosie go
Copy suites using the right-click menu
Run suites using the play button


Tutorial 2


Written by Luke Abraham 2017