v16 coupled tests can fail when output type is the tiled grid #566
Comments
|
Is this still limited to Gaea or can occur in other systems as well ? |
|
We have only seen this issue on Gaea at this point. |
|
@DeniseWorthen Do we still have issue on gaea? |
|
@junwang-noaa We turned off tiled output for this test on all platforms so it is not known whether the issue has been resolved. I suspect it is still an issue since the underlying problem was most likely due to the version of PIO we are using. I believe updating PIO is an active issue w/ the ESMF developers. Perhaps @rsdunlapiv could give an update. |
|
The 8.3 release of ESMF should have an updated PIO (to version 2.5.6). We should re-test at that point to see if the issue is resolved. |
|
moving to icebox till the 8.3 release of esmf is available for further testing |
|
@DeniseWorthen The ESMF 830bs9 is now used in the UFS WM, can you check if the issue is resolved? Thanks |
|
@DeniseWorthen @junwang-noaa 830bs09 does not have the PIO update. I suggest waiting until the ESMF 8.3.0 release to test this. Alternatively, you can test now with 830bs14 if would like. |
|
I can confirm that 830bs09 does not resolve the tiled grid output issue for the bmark_p8 test. I still get the errror |
|
I've been able to run the |
…leanup (ufs-community#566) * update submodule pointer to Man's ccpp-physics PR#944 branch: ccppv6 doc update
…ecast files (#566) This PR enables running of only the SRW App's deterministic verification (vx) tasks on staged forecast files from previous runs of the App. It partially resolves Issue #565 (it resolves the issue for deterministic vx but not ensemble vx). Specific changes: * Update lua module file for vx tasks to suppress "Logging error" messages in vx task log files. * Rename experiment variable MODEL to VX_FCST_MODEL_NAME to clarify that this is the name of the forecast model in the context of verification (and which will be used in the vx output files). This requires updates to most (all?) of the * METplus configuration files and the verification ex-scripts. * Create the new variable VX_FCST_INPUT_BASEDIR to allow the user to specify a directory in which to look for staged forecast output (instead of running a forecast). * Modify the rocoto template xml (FV3LAM_wflow.xml) to make dependencies of vx tasks on post-processing tasks appear only when the post tasks are enabled. * Add a new WE2E test category subdirectory named verification in which to group all vx tests (since more vx tasks will be coming in future PRs). Move the two existing tests MET_verification and MET_ensemble_verification from wflow_features to verification, and add a new test named MET_verification_only_vxto test the capability that this PR introduces. Note: The new WE2E test MET_verification_only_vx requires new data, specifically post-processed forecast output from the SRW App. This data needs to be staged on each platform; currently, it is located in a personal directory on Hera.
PR #566 changed the variable "MODEL" to a more descriptive name, but failed to make this change in config.community.yaml. The unit tests for generate_FV3LAM_wflow.py make use of this file as an input config.yaml, so they are now failing due to this incorrect variable name. This wasn't caught because prior to #558 the unit tests were broken for a different reason. This change simply makes the appropriate rename, which should fix the failing unit test. Also created an f-string that was missed in a setup.py error message.
Description
This issue was first noted on Gaea with the coupled V16 tests failing with a PIO error.
To Reproduce:
Run the cpld_bmarkfrac_v16 test on Gaea using tiled grid output.
Additional context
The initial issue was detailed in Issue #509. It was shown that:
The issue appears to be related to the version and/or integration of PIO within ESMF.
The text was updated successfully, but these errors were encountered: