Comments (8)
More information from @hguo-gfdl. Some is included above, but this helps clarify the issue:
Email from @hguo-gfdl
I ran into problematic regional output "vmo" (Ocean Mass Y Transport) along Carribean Windward Passage after changing ocean layout. Originally I used around 8100 PEs as below,
<freInclude name="resourceparams_A1152x2_O5791_production">
<resources jobWallclock="16:00:00" segRuntime="2:30:00">
<atm ranks="1152" threads="2" layout = "8,24" io_layout = "1,4" />
<lnd layout = "8,24" io_layout = "1,4" />
<ice layout = "144,8" io_layout = "1,4" />
<ocn ranks="5791" threads="1" layout = "90,90" io_layout = "1,5" mask_table="mask_table.2309.90x90"/>
</resources>
</freInclude>
After reducing to 6408 PEs, "vmo" at Windward Passage reached extremely large value, which was unreasonable,
<freInclude name="resourceparams_A864x2_O4671_production">
<resources jobWallclock="16:00:00" segRuntime="2:40:00">
<atm ranks="864" threads="2" layout = "6,24" io_layout = "1,4" />
<lnd layout = "6,24" io_layout = "1,4" />
<ice layout = "72,12" io_layout = "1,4" />
<ocn ranks="4671" threads="1" layout = "90,72" io_layout = "1,4" mask_table="mask_table.1809.90x72"/>
</resources>
</freInclude>
The xml for the above run is on gaea:
/ncrc/home2/oar.gfdl.cmip6/xml/CM4/CM4Bling_am4p0c96L33_OM4p25_piControl_C.xml
The untarred history file is at gfdl:
/archive/oar.gfdl.cmip6/CM4/warsaw_201710_om4_v1.0.1/CM4_piControl_C/gfdl.ncrc4-intel16-prod-openmp/history/tmp_650/06500101.ocean_Windward_Passage.nc
The stdout is in gfdl:
/archive/oar.gfdl.cmip6/CM4/warsaw_201710_om4_v1.0.1/CM4_piControl_C/gfdl.ncrc4-intel16-prod-openmp/ascii/06500101.ascii_out.tar
from fms.
@hguo-gfdl could you please point me to a file that has the correct (or what looks correct) for the vmo
variable? I grabbed two random files from /archive/oar.gfdl.cmip6/CM4/warsaw_201710_om4_v1.0.1/CM4_piControl_C/gfdl.ncrc4-intel16-prod-openmp/history
and see no difference. It would be easier if you could point me to two versions of the variable: one with the expected output, and the one with the bad output.
from fms.
@underwoo The expected output:
gfdl:/archive/oar.gfdl.cmip6/CM4/warsaw_201710_om4_v1.0.1/CM4_piControl_C/gfdl.ncrc4-intel16-prod-openmp/history/tmp_293/02930101.ocean_Windward_Passage.nc
The bad output:
gfdl:/archive/oar.gfdl.cmip6/CM4/warsaw_201710_om4_v1.0.1/CM4_piControl_C/gfdl.ncrc4-intel16-prod-openmp/history/tmp_650/06500101.ocean_Windward_Passage.nc
from fms.
from fms.
@underwoo I just added read permission on these files. Please let me know if there are any problems.
Thanks,
from fms.
@hguo-gfdl this might be related to another issue: #52. This issue points to an inconsistency in how diag_manager_mod
names axes. I will need to run a test case, both to see if I can reproduce the error and what information is in the uncombined history files. Hopefully I will have some information next week.
from fms.
@hguo-gfdl I was able to reproduce the issue, and it does appear to be a diag_manager issue, more than likely this is not related to what we see in #52.
In your case, with this layout, diag_manager produces two files (since this is a region). One of those files has the variables volcello
, thetao
and so
. The other file has variable vmo
and vo
. Uncombined, the variable outputs look correct. When combined, only the variables vmo
and vo
exist in the combined file, but have the data for volcello
and thetao
respectively.
We will continue to look into this issue, and try to come up with a fix.
Thank you for reporting this.
from fms.
We think we have discovered the issue, and it is something we do not know how to fix at this time. We believe what is happening is that at different resolutions, some of the MPI processes may not have access to all the same variables within small regions due to different grids used in the model. The regional output process writes a file per MPI process, and simply dumps the data to the file. If the variable doesn't exist within the grid on a particular MPI process, the variables is unregistered on that process. Note, this particular case will only happen with certain variables on the ocean, ice and land component grids.
We will update, in a future patch, the workflow to not combine files that do not have the same variables set of variables. This will keep the workflow from overwriting data. However, it is not a fix to this issue.
If we discover a method to correct this behavior, we will let you know. For now, I will close this as something we will not fix.
from fms.
Related Issues (20)
- data_override_0d bug
- Expand data_override unit tests
- configure fails on gaea
- ifort warning in fms_string_utils HOT 8
- missing registered fields in diag_field_log.out HOT 7
- Modern_diag_manager: 0 days output frequency not working
- Modern_diag_manager :: deprecate `mix_snapshot_average_fields`in the new diag manager
- Modern_diag_manager:: `flush_nc_files` is not implemented in the new diag manager HOT 1
- Modern_diag_manager:: Crashes when two+ empty files
- `test_mpp_gatscat` causes compilation error with oneapi 2024.1 HOT 1
- Attempt to write out u_atm, etc. fails when using a mask_table HOT 8
- Modern diag_manager: The `weight` argument is not passed into reduction method functions
- Consistent maximum length of filenames? HOT 1
- Update the field_table yaml format
- Update the data_table yaml format
- Modern diag manager: Wrong time and time_bnds when only writing 1 time dimension
- Modern diag_manger: Crashes if freq is 0 days and you have static fields
- Diag manager: Allow scalar static variables
- FMS does not appear to compile with GFortran 14.1 HOT 3
- Modern diag_manager:: diag_send_complete_instant has not been implemented HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from fms.