github actions utest

[MinsukJi-NOAA]

This PR addresses issue #168

• Include continuous integration related files (scripts, github actions yaml file, dockerfile, etc.)
• Include updates to the utest script

[MinsukJi-NOAA]

RT on Orion keeps failing, due to compile_13 failing with this error message:

/work/noaa/stmp/jminsuk/stmp/jminsuk/FV3_RT/rt_420215/compile_13/build_fv3_13/FV3/ccpp/physics/ccpp_static_api.F90(966): catastrophic error: **Internal compiler error: internal abort** Please report this error along with the circumstances in which it occurred in a Software Problem Report. Note: File and line given may not be explicit cause of this error. compilation aborted for /work/noaa/stmp/jminsuk/stmp/jminsuk/FV3_RT/rt_420215/compile_13/build_fv3_13/FV3/ccpp/physics/ccpp_static_api.F90 (code 1)`

[climbfuji]

RT on Orion keeps failing, due to compile_13 failing with this error message:

/work/noaa/stmp/jminsuk/stmp/jminsuk/FV3_RT/rt_420215/compile_13/build_fv3_13/FV3/ccpp/physics/ccpp_static_api.F90(966): catastrophic error: **Internal compiler error: internal abort** Please report this error along with the circumstances in which it occurred in a Software Problem Report. Note: File and line given may not be explicit cause of this error. compilation aborted for /work/noaa/stmp/jminsuk/stmp/jminsuk/FV3_RT/rt_420215/compile_13/build_fv3_13/FV3/ccpp/physics/ccpp_static_api.F90 (code 1)`

Hmm. I have seen these internal compiler errors for other projects in the past, it often had to do with a too complicated optimization task that the compiler is trying to accomplish. We need to figure out which COMPILE command build_13 corresponds to, and what line the compiler is complaining about. Which version of the Intel compiler are you using?

[MinsukJi-NOAA]

Hmm. I have seen these internal compiler errors for other projects in the past, it often had to do with a too complicated optimization task that the compiler is trying to accomplish. We need to figure out which COMPILE command build_13 corresponds to, and what line the compiler is complaining about. Which version of the Intel compiler are you using?

modulefiles/orion.intel/fv3 module loads intel/2018

compile 13 appears to be CCPP=Y and DEBUG=Y

[climbfuji]

Hmm. I have seen these internal compiler errors for other projects in the past, it often had to do with a too complicated optimization task that the compiler is trying to accomplish. We need to figure out which COMPILE command build_13 corresponds to, and what line the compiler is complaining about. Which version of the Intel compiler are you using?

modulefiles/orion.intel/fv3 module loads intel/2018

compile 13 appears to be CCPP=Y and DEBUG=Y

Interesting. I recently made this change (and it worked on all machines, including orion) so that we have at least one compile command that tests compiling the code without providing any suites (i.e. compile all available suites). What you can do is to duplicate this line N times (where N is the number of machines, i.e. one line for hera.intel, one for cheyenne.intel, ...) and modify the line for orion to include just the suites that are required to run the following tests.

[MinsukJi-NOAA]

Interesting. I recently made this change (and it worked on all machines, including orion) so that we have at least one compile command that tests compiling the code without providing any suites (i.e. compile all available suites). What you can do is to duplicate this line N times (where N is the number of machines, i.e. one line for hera.intel, one for cheyenne.intel, ...) and modify the line for orion to include just the suites that are required to run the following tests.

That change allowed the RT to pass. Thanks.

[MinsukJi-NOAA]

Preliminary CI documentation can be found here: CI tests for UFS-weather-model

[MinsukJi-NOAA]

For an example of a CI test, see Add hera RT results

For an example of a skipped CI test, see Add orion RT results

[junwang-noaa]

@DusanJovic-NOAA I remember Dom said the github action running time for public repo is unlimited, also because of the characteristics of weather model forecast tests, we have to run jobs longer than a few minutes, so what are the issues?

…

On Thu, Sep 3, 2020 at 10:59 AM Dusan Jovic ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In tests/ci/json_helper.py <#169 (comment)> : > @@ -0,0 +1,46 @@ +#!/usr/bin/env python3 Exactly. That's why I think these kind of tests (tests that run longer than few minutes) are simply not suitable for these "continuous integration" testing environments. — You are receiving this because your review was requested. Reply to this email directly, view it on GitHub <#169 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI7D6TML6CPXCJLC6HASFQTSD6VMFANCNFSM4O4Z2ASA> .

[climbfuji]

@DusanJovic-NOAA I remember Dom said the github action running time for public repo is unlimited, also because of the characteristics of weather model forecast tests, we have to run jobs longer than a few minutes, so what are the issues?
…

Have a look at https://github.com/pricing, scroll down to "Compare features" please.

[MinsukJi-NOAA]

Updated utest documentation can be found here: Unit Test for UFS-weather-model

[MinsukJi-NOAA]

All regression tests passed on Hera, Orion, and Dell. All unit tests except restart test passed on Hera, Orion, and Dell. This PR is ready for merge.

[climbfuji]

Looks ok to me and my limited understanding of the CI design.