New diagnostic nc writer: error when writing surface variables

[akshaysridhar]

In the branch https://github.com/CliMA/ClimaAtmos.jl/tree/as/test-viz-newdiag, I'm attempting to test the new diagnostic output specifications. I'm using this file to test the output workflow. When specifying the h5 writer, I get successful simulation outcomes (see truncated trace below) with generated files e.g. cli_6.0h_average_21600.0.h5 for the requested variables/intervals.

┌ Info: Running
│   job_id = "sphere_aquaplanet_rhoe_equilmoist_allsky_gw_res"
│   output_dir = "output/sphere_aquaplanet_rhoe_equilmoist_allsky_gw_res"
└   tspan = (0.0f0, 43200.0f0)
[ Info: solve!: 188.394 s (282638063 allocations: 119.16 GiB)
[ Info: sypd: 0.6282364633899803
....
....
┌ Warning: Buildkite not detected. Skipping regression tests.
└ @ Main ~/Research/Codes/ClimaAtmos.jl/regression_tests/compute_mse.jl:58
[ Info: Please review output results before merging.
[ Info: Callback verification, n_expected_calls: AbstractFloat[0.6666666666666666, 2.0f0, 108.0f0]
[ Info: Callback verification, n_measured_calls: [[1], [3], [108]]

When switching the output writer to nc however, this simulation fails - seemingly due to specific output keys (I assume invalid keys would be caught even in the h5 writer), since I still see files like

@Sbozzolo Can you help check this out please? Thanks! The first diag variable at which a run with this config file above fails is ts, or in its absence, hfes- both are surface vars on the sphere.

[Sbozzolo]

Is this with the main branch?

[akshaysridhar]

Is this with the main branch?

Working branch is https://github.com/CliMA/ClimaAtmos.jl/tree/as/test-viz-newdiag but main has been merged into this as of commit caded4d1

[akshaysridhar]

I suspect it's something to do with the surface variables

[Sbozzolo]

Okay, the NetCDF writer you are currently does not work with surfaces (and many other cases), but that has been fixed in this branch.

I working out some issues with performance and GPUs and then we can merge it

[akshaysridhar]

Noted thanks: Pulling the branch associated with #2179 > will report results shortly.

[akshaysridhar]

Works in commit 49105. @Sbozzolo Thanks! Resolved, closing issue.