search

NOAA-EMC / RDASApp

Regional DAS
GNU Lesser General Public License v2.1
1 stars 13 forks source link

update RDASApp submodules to match mpasjedi-v3.0.1 #194

Closed Junjun-NOAA closed 1 month ago

Junjun-NOAA commented 1 month ago

List of submodule changes. issue #193 

ioda             c7b8760f -> d49ed17e
ufo              92ccfb2a -> 94d50d64
oops             35820130 -> d77217323
vader            e3457cba -> 6d56a1eb5
mpas             3ecd59e2 -> 41e9a3fb8   #  repo URL also changed
mpas-jedi        a1c60997 -> b9d596d7c
#fv3-jedi         d3c800b8 -> c99519638
#fv3-jedi-lm      a6e97d76 -> 30ef7a390
SamuelDegelia-NOAA commented 1 month ago

Those ctest times include waiting time in the queue. Hera has been quite slow for me recently so I am guessing that is the problem. Typically the longest ctest is the GETKF solver which takes about ~20 minutes of actual runtime (it will be reduced to ~10 min after we can use reduce obs space).

SamuelDegelia-NOAA commented 1 month ago

Part of the increased wait time too is probably due to requesting 30 min for the mpas-jedi ctests. The increased wall time is really only needed for the GETKF solver so I plan to tune those a bit better whenever I update the OOPS hash (after this PR is merged).

guoqing-noaa commented 1 month ago

ctest takes 35919.88 sec. Is it an issue related to HPC?

@ShunLiu-NOAA Yes, as @SamuelDegelia-NOAA mentioned, this was because Hera was known to have a very long job waiting time recently. The ctests themselves don't have any issues.

guoqing-noaa commented 1 month ago

Part of the increased wait time too is probably due to requesting 30 min for the mpas-jedi ctests. The increased wall time is really only needed for the GETKF solver so I plan to tune those a bit better whenever I update the OOPS hash (after this PR is merged).

Thanks for the thought. I don't think 10 minutes vs 30 minutes will make any evident differences in RDHPCS, :)

SamuelDegelia-NOAA commented 1 month ago

Part of the increased wait time too is probably due to requesting 30 min for the mpas-jedi ctests. The increased wall time is really only needed for the GETKF solver so I plan to tune those a bit better whenever I update the OOPS hash (after this PR is merged).

Thanks for the thought. I don't think 10 minutes vs 30 minutes will make any evident differences in RDHPCS, :)

I wouldn't think so, but 5 min vs 30 min is currently making a big difference. The fv3-jedi tests that request 5 min here are not waiting at all and run within a few minutes. But the mpas-jedi tests that request 30 min are waiting much much longer.

guoqing-noaa commented 1 month ago

Part of the increased wait time too is probably due to requesting 30 min for the mpas-jedi ctests. The increased wall time is really only needed for the GETKF solver so I plan to tune those a bit better whenever I update the OOPS hash (after this PR is merged).

Thanks for the thought. I don't think 10 minutes vs 30 minutes will make any evident differences in RDHPCS, :)

I wouldn't think so, but 5 min vs 30 min is currently making a big difference. The fv3-jedi tests that request 5 min here are not waiting at all and run within a few minutes. But the mpas-jedi tests that request 30 min are waiting much much longer.

Good to know. If it helps, you may go ahead to make the changes. Thanks.

SamuelDegelia-NOAA commented 1 month ago

Good to know. If it helps, you may go ahead to make the changes. Thanks.

I plan to wait until we update the OOPS hash further. We'll need to do that first to get the lower run times for GETKF (via reduce obs space). Hopefully I can start on that as soon as this PR gets merged.

ShunLiu-NOAA commented 1 month ago

@hu5970 Do you get a chance to review this PR?

guoqing-noaa commented 1 month ago

@ShunLiu-NOAA @hu5970 Do we still lingering issues on this PR? Otherwise, we should consider merging this PR. Thanks!

hu5970 commented 1 month ago

@guoqing-noaa I will ask the team on this PR tomorrow and try to arrange an order to merge the PRs.