Issue with Internal Tide Test Case – Velocity Field Returns NaN (internal_tide.jl)

[kravi1220]

Hi,

I was running the internal tide test case, and while it initially progressed for a few time steps, the velocity field starts returning NaN values, and the model subsequently stops.

Could you please take a look at the code, or let me know if I might be missing something in the setup?

Best regards,
Ravi

[glwagner]

What version of Julia are you using? The tests run with 1.10.

[kravi1220]

I am using Version 1.11.5

[kravi1220]

I am planning to conduct an internal tide idealized experiment using my own set of slice sections and bathymetry, based on the provided test case. I would like to confirm whether version 1.10 of the software is compatible with the updated Oceananigans/ClimaOcean options for running these experiments.

Could you please advise if this version supports the necessary configurations, or if a different version is recommended?

[kravi1220]

Still same issue with 1.10 too

[glwagner]

The documentation is built on ubuntu with julia 1.10.9 and runs successfully: https://clima.github.io/OceananigansDocumentation/stable/literated/internal_tide/

I don't think we have knowledge of why a version (or machine) would fail, I am sorry I cannot give you that information right now. Can you tell us what machine you are using?

Could you please take a look at the code, or let me know if I might be missing something in the setup?

Are you running the internal tide test case exactly as it appears, or something else / something changed?

[kravi1220]

I am using Ubuntu 24.04.2 LTS. I tried most of version of v1.10.* and all version having same issue

[glwagner]

There are reports of the internal tide test case blowing up on 1.11 and ubuntu (but succeeding on Mac / ARM) here: #3836

I just merged main onto that branch so we will see what it does.

I am surprised that it fails on 1.10.9 when the test succeed. However, you are running a later version of ubuntu than the CI server (24.04.2 LTS vs 20.04.6 LTS). It is possible that can explain the discrepancy?

[kravi1220]

Somehow I manage to work the test case with version 1.10.9. but when switching bathymetry from test case to real data again it producing NaN after few iteration.

[kravi1220]

I’ve resolved the issue, and it’s now working with version 1.11.5 as well. Apologies — the problem with my experiment was due to using a very high horizontal resolution without adjusting the time step (dt) accordingly.

[glwagner]

ah ok, I didn't comprehend that you had changed the script!

Feel free for future issues to post your entire script. We don't mind lots of code! You don't need to post your script if it is identical to the example script, but any change at all (as you have discovered) can cause a change in behavior. So the only solution is to read all the code, to figure out what's wrong.

I'll close this then, happy that your issue is resolved.