Opened 10 years ago

Closed 8 years ago

#456 closed help (fixed)

UMUI and start up files

Reported by: mtjb25 Owned by: lois
Component: UM Model Keywords:
Cc: w.mcginty@… Platform:
UM Version: 7.6

Description

Hi,

I am doing my first run of the UM. I have an account on Hector, and I am using the UMUI on PUMA. I am a little uncertain about some of the required fields for UMUI, the HELP and other documentation seem to refer to running UM at the Met Office.

In the general details tab I have:
Target machine (my hector userid)
mail id (my email address)
jobname (RUN1D000)

After hitting submit I get another window which says my target machine is IBM and hostname is hpc1e
Define submission method (Loadleveler for IBM platform)
Class name (parallel)
job time limit (1000)
job memory limit (10)
maximum compilation processes (3)

I don't know if these settings are correct and if not I don't know what to change.

Also I have requested startfiles for the dates
24/6/2007 - 26/6/2007,
30/6/2008 - 2/7/2008,
6/6/2008 - 8/7/2008,

I did this by replying to an earlier answer, so I don't know if my request was recieved. Will this data allow me to run the global set up?

Thank you

Michaela

Change History (9)

comment:1 Changed 10 years ago by willie

  • Owner changed from um_support to willie
  • Status changed from new to assigned

Hi Michaela,

Your basis file tcysa is empty. What is it that you are trying to do? It is best to start from a basis file that is known to work and then modify it step by step. There are lots of standard jobs under UMUI user 'umui'

Regards,

Willie

comment:2 follow-up: Changed 10 years ago by lois

Would it be useful Michaela to have a Unified Model 'driving lesson'? There are slides from previous one day courses on the CMS web pages as well as lots of documentation but we would be very happy to visit you or invite you to Reading for a day to help you start out using the UM.

As Willie says you really need to have a UM job that is either a standard job or perhaps from the Met Office or from a colleague to start from.

Lois

comment:3 in reply to: ↑ 2 Changed 10 years ago by mtjb25

Thank you for your response.
I downloaded the basis file from the met office site: http://collab.metoffice.gov.uk/twiki/bin/view/Support/ParallelSuite24

it is in umui_jobs/basis_qayla.

I then loaded the basis file as per the instructions
http://ncas-cms.nerc.ac.uk/index.php/um-faq/45-loading-umui-job-basis-file[[BR]]

so I don't know why it's empty, I can repeat the process again.

I attended the UM users course at the Met Office, which was very useful. Since we were running the UM on met offices machines, I didn't have to change too many settings regarding target machine etc since it was already set up for use. Since I am now trying to run the UM on a machine not on the Met Office site I think I will have to change some of these general settings. I have looked at other people's jobs to try to work out what these might be, but there is no consistency, so I remain unsure.
I have also viewed

http://ncas-cms.nerc.ac.uk/index.php/component/content/article/67-um-support/1478-tutorial-umui-intro[[BR]]

but I can't find an expample of the UMUI window I get after submitting the general information.

Would you like me to print the screen so you have a better idea of the information I am asking for?

The aim of the work I am involved is part of the Flood Risk Research Consortium
http://www.floodrisk.org.uk/[[BR]]

I working with Brian Golding on the following work package:

'To extend the development of the new state-of-the-art STEPS and MOGREPS systems developed at the Met. Office and apply them to the problem of urban flooding on small space scales, with lead times of 2-3 hours in extreme rainfall conditions.'

To date I have been experimenting with WRF and MM5 since these are readily available, easily accessed and straight forward to run. However after consulting with Brian Golding and our FRMRC partners I am now going to use the UM model since this is the model which is operational in the UK. The initial plan is to run the UM for some case studies which have been chosen by our partners Sheffield University and Imperial College, who will use the model output to drive their urban flood models.

Is the data for the start dates I requested available? If not I can use ECMWF data, I know the ECMWF data needs to be adjusted before it can be used with the UM, is there any documentation on how to do this?

Thank you for your help

Michaela

Replying to lois:

Would it be useful Michaela to have a Unified Model 'driving lesson'? There are slides from previous one day courses on the CMS web pages as well as lots of documentation but we would be very happy to visit you or invite you to Reading for a day to help you start out using the UM.

As Willie says you really need to have a UM job that is either a standard job or perhaps from the Met Office or from a colleague to start from.

Lois

comment:4 Changed 10 years ago by lois

  • Cc w.mcginty@… added
  • Owner changed from willie to lois

Thanks for the information Michaela.

There are a significant number of changes that need to be made when uploading a basis file from the collaboration wiki. We do not have information on our web pages yet with all the steps that are needed for this and I imagine this information will change with different jobs. You are the first to try this!

The changes will not only be for your userid and mail id but will need to be

  • for the target machine, type, job submission, TIC code
  • FCM, including all extras that are mentioned on the collaboration wiki page
  • hand edits and STASHmaster files on PUMA
  • directories on HECToR

I will go through an example on Monday with Ros Hatcher (our FCM manager) and we will write up our experience to form some new web pages.

Willie McGinty? should be able to get you the start files you need either from the Met Office or ECMWF. If they are from ECMWF he should be able to set these up for you to be used as UM start files. I will leave Willie to contact you when the files are available.

Lois


comment:5 Changed 10 years ago by lois

Hello Michaela

We have been reviewing all the PS24 UM jobs on the collaboration wiki and after our discussion we consider that it would be a very challenging job for you, as a newcomer to the UM, to undertake the transfer of these operational jobs to a new computer. It is not just a question of just changing 1 or 2 things I have two pages of notes from just the global compile job!

Our suggestion is

  • we undertake the transfer of the PS24 jobs to HECToR and make notes of all the changes and decisions that have to be made to feed back to the Met Office who firstly may be able to change how they package all these operational jobs on the collaboration wiki and secondly make these notes available to users who then want to try to do this for other paralell suite jobs.
  • while Willie is undertaking this work, which he says may take 2-3 weeks for all the jobs (down to the variable resolution version) in the parallel suite, you could be getting used to the UM and the academic computer service HECToR using PS22 set of jobs which are already available on the PUMA UMUI. To look at these jobs, already set up for use on HECToR, you need to search the PUMA UMUI using the filter umui. The userid umui is where we put all our standard jobs for weather and for climate. These jobs can be copied and the ony change you need to make to run them is your userid, email and TIC code on HECToR. They should be relatively robust and will allow you to try running through the whole suite and then working out what output you want and need and then how to analyse and visualise the output data.
  • Willie is collecting together the start data you requested and then he will sort out any issues of starting from ECMWF data. This involves a bit of non-standard Met Office code which needs testing for the latest version of the model which PS24 uses. We have only just installed the lastest version of the UM on HECToR.

Let us know if this is an acceptable plan

Lois

comment:6 Changed 10 years ago by willie

Hi Michaela,

Just a progress report. I have got the global build and reconfiguration jobs to run on HECToR. So the required branches have been installed, and the ancillary trees transferred (all 220Gbytes!). However, the run job crashes with a segmentation fault:

Fatal error in MPI_Recv: Invalid count, error stack:
MPI_Recv(186): MPI_Recv(buf=0x7fffd666afb0, count=-288, dtype=USER<struct>, src=3, tag=10, comm=0x84000003, status=0x7fffcf298450) failed
MPI_Recv(104): Negative count, value is -288
aborting job:

I am working to resolve this issue and will keep you posted.

Regards,

Willie

comment:7 Changed 10 years ago by willie

  • UM Version changed from <select version> to 7.6

Hi Michaela,

I have not been able to fix the segmentation fault. It has been suggested that I move to UM7.6 as this is scientifically the same. However, I have not been able to get 7.6 working either (another segmentation fault). I could do a manual merge of the conflicting patches in the UM7.6 code, but I am not au fait with the code and might make poor choices. An alternative is to wait a couple of weeks until UM7.7 is out and hope that when transferred to HECToR the problem will be solved. Another possibility is for you to work with PS22 (UM7.3) in the interim as this is working on HECToR - see the standard umui jobs under experiment xeri.

Apologies for the delay in getting this working.

Regards,

Willie

comment:8 Changed 10 years ago by willie

Hi Michaela,

I have now got the PS24 Global run to work. This required a new patch from the Met Office and some experimentation on my part. I have made a set of standard jobs xfkc under user name umui. Let me know how you get on with this. I'll now start setting up the other sub domains.

Again, apologies for the delay in getting this to work.

Regards,

Willie

comment:9 Changed 8 years ago by ros

  • Component changed from UMUI to UM Model
  • Resolution set to fixed
  • Status changed from assigned to closed
Note: See TracTickets for help on using tickets.