More mira port #258

jedwards4b · 2016-07-14T17:37:11Z

This change gets a basic case setup working on mira again. However the workflow - running st_archive and then resubmitting does not yet work. This means that ERR and ERI tests will not work. mpi-serial is also not supported on this platform so most of the scripts_regression_tests will also fail.

Also including updates (machine file updates) for machines rarely used/tested - Updating batch system type for babbage, brutus, eos, erebus, hera, pleiades-wes, sierra, titan.

Also build output is now logged, streaming it to a file, instead of buffering it in memory.

Test suite:
Test baseline:
Test namelist changes:
Test status: [bit for bit, roundoff, climate changing]

Closes #245
User interface changes?:

Code review: jayesh

jayeshkrishna · 2016-07-14T17:39:36Z

I am going to try testing this change using CIME_MODEL=cesm with the following case,

./create_newcase -case cime_sanity_check_mira -compset X -res f19_g16

jayeshkrishna · 2016-07-14T17:41:04Z

This PR also seems to have changes for other (not mira) machines. Is that intentional?

jedwards4b · 2016-07-14T17:43:38Z

cime_config/cesm/machines/config_batch.xml

@@ -142,21 +143,21 @@
   </batch_system>

  <!-- babbage is PBS -->
-  <batch_system MACH="babbage" version="x.y">


I updated some config_batch.xml for machines that are rarely used or tested.

jayeshkrishna · 2016-07-14T19:37:23Z

@jedwards4b : I updated the PR message with the information (changes not related to mira) above. Please go through it and feel free to update/modify the new comments.

jedwards4b · 2016-07-14T20:06:47Z

looks good, thank you

jayeshkrishna · 2016-07-14T21:02:03Z

I have submitted a job on Cetus for CIME_MODEL=cesm (create/setup/build/submit succeeds).

The ./create_newcase command failed with CIME_MODEL=acme . This will need to be pursued as part of Issue #245 . The output from create_newcase (CIME_MODEL=acme) is shown below,

[jayesh@miralac1 scripts (master>)]$ ./create_newcase -case acme_sanity_check_mira -compset X -res f19_g16                       
Compset longname is 2000_XATM_XLND_XICE_XOCN_XROF_XGLC_XWAV 
Compset specification file is /gpfs/mira-home/jayesh/acme/ESMCI_cime_merge/scripts/Tools/../../driver_cpl/cime_config/config_compsets.xml
Pes     specification file is /gpfs/mira-home/jayesh/acme/ESMCI_cime_merge/scripts/Tools/../../driver_cpl/cime_config/config_pes.xml
Pes setting: grid          is a%1.9x2.5_l%1.9x2.5_oi%gx1v6_r%r05_m%gx1v6_g%null_w%null 
Pes setting: compset       is 2000_XATM_XLND_XICE_XOCN_XROF_XGLC_XWAV 
Pes setting: grid match    is any 
Pes setting: machine match is mira 
Pes setting: compset_match is [DX]ATM 
Pes setting: pesize match  is any 
ERROR: Command: '/usr/bin/xmllint --format --output /gpfs/mira-home/jayesh/acme/ESMCI_cime_merge/scripts/acme_sanity_check_mira/env_mach_pes.xml -' failed. See terminal output

jedwards4b · 2016-07-14T21:12:30Z

The PR has been updated with a fix for the above, please try again.

jayeshkrishna · 2016-07-15T01:40:12Z

Got the following error with CIME_MODEL=cesm (in the job error output),

Not able to fully resolve item '/usr/bin/runjob   $LOCARGS  --envs OMP_STACKSIZE=64M  --envs OMP_NUM_THREADS=8  --envs BG_THREADLAYOUT=1  --label short  -p 8  -n 8 :  /projects/ClimateEnergy/usr/jayesh/cesm_sanity_check_mira/bld/acme.exe  >> acme.log.160714-213314 2>&1 '
ERROR: Command: '/usr/bin/runjob   $LOCARGS  --envs OMP_STACKSIZE=64M  --envs OMP_NUM_THREADS=8  --envs BG_THREADLAYOUT=1  --label short  -p 8  -n 8 :  /projects/ClimateEnergy/usr/jayesh/cesm_sanity_check_mira/bld/acme.exe  >> acme.log.160714-213314 2>&1 ' failed from dir '/projects/ClimateEnergy/usr/jayesh/cesm_sanity_check_mira/run'. See terminal output

@jedwards4b : Can you build and run any case on Mira/Cetus with CIME_MODEL=cesm ? I could try the case that you are using in your tests to see if the fix works.

jayeshkrishna · 2016-07-19T20:58:39Z

I have submitted a job (case creates+builds+submits correctly) on Mira for CIME_MODEL=cesm .
This PR still does not fix issues for CIME_MODEL=acme. I can work on that since CIME_MODEL=cesm is working now.

jedwards4b · 2016-07-19T21:08:07Z

I'm sorry I thought that it would work for acme - what problem are you seeing?

jayeshkrishna · 2016-07-20T13:56:34Z

No worries, I will start looking at the issues (the first one was due to missing PES_PER_NODE tag). Let me try to get it working and will let you know the details.

jayeshkrishna · 2016-07-20T15:47:35Z

I got the following error at runtime (CIME_MODEL=cesm) after merging the branch locally to master,

Not able to fully resolve item '/usr/bin/runjob   $LOCARGS  --envs OMP_STACKSIZE=64M  --envs OMP_NUM_THREADS=8  --envs BG_THREADLAYOUT=1  --label short  -p 8  -n 8 :  /projects/ClimateEnergy/usr/jayesh/cesm_sanity_check_mira/bld/acme.exe  >> acme.log.160719-210632 2>&1 '
ERROR: Command: '/usr/bin/runjob   $LOCARGS  --envs OMP_STACKSIZE=64M  --envs OMP_NUM_THREADS=8  --envs BG_THREADLAYOUT=1  --label short  -p 8  -n 8 :  /projects/ClimateEnergy/usr/jayesh/cesm_sanity_check_mira/bld/acme.exe  >> acme.log.160719-210632 2>&1 ' failed from dir '/projects/ClimateEnergy/usr/jayesh/cesm_sanity_check_mira/run'. See terminal output

Will try it with master and see how it goes.

jedwards4b · 2016-07-20T16:05:34Z

The "not able to fully resolve" is not an error, it is expected on mira because of the LOCARGS flag.
Look in acme.log.160719-210632 for the real problem.

jayeshkrishna · 2016-07-20T16:15:00Z

The error message above is with CIME_MODEL=cesm, and I only have JOBID.{output|error|cobaltlog} in the case directory
(PS: Meanwhile, I built it on Mira and ran it on Cetus - not sure if that makes a difference)

jedwards4b · 2016-07-20T16:16:49Z

the acme.log file will be in the run directory

jayeshkrishna · 2016-07-20T16:27:44Z

Looks like an issue in my environment, although I had set CIME_MODEL=cesm while building the model my shell startup files had CIME_MODEL=acme. After changing it to CIME_MODEL=cesm, the job ran successfully.
Going to try the ACME side now.

rljacob · 2016-07-20T22:07:49Z

Did this also fix #253 ?

jedwards4b · 2016-07-20T22:08:58Z

I believe so

Several changes to get Mira working was added in 1f9f0b3 . However some build and runtime issues remained with CIME_MODEL=acme . This commit adds the remaining fixes required to get the code working on Mira for CIME_MODEL=acme. * Fixed the batch submit command * Added the PES_PER_NODE tag * Fixed the name of the default project (now called HiRes_EarthSys_2) * Modified the locargs argument for runjob (--block) These fixes are already present in cesm (CIME_MODEL=cesm). Also see Issue #245 and PR #258

This simply uses github's suggested code of conduct, which looks good to me. Fixes ESMCI#258

jedwards4b added 2 commits July 13, 2016 15:48

Merge remote-tracking branch 'mydev/test_updates2'

cd50e4f

Merge branch 'master' of https://github.com/ESMCI/cime

0f298cf

jedwards4b added the in progress label Jul 14, 2016

jedwards4b assigned jayeshkrishna Jul 14, 2016

jedwards4b reviewed Jul 14, 2016
View reviewed changes

jayeshkrishna assigned jedwards4b Jul 15, 2016

jayeshkrishna mentioned this pull request Jul 15, 2016

Update config_batch.xml #257

Closed

jedwards4b added 9 commits July 18, 2016 15:14

Merge branch 'master' of https://github.com/ESMCI/cime

557cff3

Merge branch 'master' of https://github.com/ESMCI/cime

706c28e

get mira working again

bfac10e

correct batch directives

8a5ef7f

update acme side

a0c984d

changed so that the expression is evald

b90daa8

more updates for mira

77b4ae1

sta working now

acb6008

get st_archive working on mira

037d253

jedwards4b force-pushed the more_mira_port branch from 17b8a33 to 037d253 Compare July 19, 2016 19:55

remove commented code

84ebf2d

jedwards4b merged commit 84ebf2d into ESMCI:master Jul 19, 2016

jedwards4b removed the in progress label Jul 19, 2016

jayeshkrishna mentioned this pull request Jul 21, 2016

Get Mira working with acme #276

Merged

jedwards4b deleted the more_mira_port branch July 26, 2016 20:32

pesieber pushed a commit to pesieber/cime that referenced this pull request Mar 15, 2023

Add a code of conduct

21ac335

This simply uses github's suggested code of conduct, which looks good to me. Fixes ESMCI#258

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

More mira port #258

More mira port #258

jedwards4b commented Jul 14, 2016 •

edited by jayeshkrishna

Loading

jayeshkrishna commented Jul 14, 2016

jayeshkrishna commented Jul 14, 2016

jedwards4b Jul 14, 2016

jayeshkrishna commented Jul 14, 2016

jedwards4b commented Jul 14, 2016

jayeshkrishna commented Jul 14, 2016

jedwards4b commented Jul 14, 2016

jayeshkrishna commented Jul 15, 2016

jayeshkrishna commented Jul 19, 2016

jedwards4b commented Jul 19, 2016

jayeshkrishna commented Jul 20, 2016

jayeshkrishna commented Jul 20, 2016

jedwards4b commented Jul 20, 2016

jayeshkrishna commented Jul 20, 2016 •

edited

Loading

jedwards4b commented Jul 20, 2016

jayeshkrishna commented Jul 20, 2016

rljacob commented Jul 20, 2016

jedwards4b commented Jul 20, 2016

More mira port #258

More mira port #258

Conversation

jedwards4b commented Jul 14, 2016 • edited by jayeshkrishna Loading

jayeshkrishna commented Jul 14, 2016

jayeshkrishna commented Jul 14, 2016

jedwards4b Jul 14, 2016

Choose a reason for hiding this comment

jayeshkrishna commented Jul 14, 2016

jedwards4b commented Jul 14, 2016

jayeshkrishna commented Jul 14, 2016

jedwards4b commented Jul 14, 2016

jayeshkrishna commented Jul 15, 2016

jayeshkrishna commented Jul 19, 2016

jedwards4b commented Jul 19, 2016

jayeshkrishna commented Jul 20, 2016

jayeshkrishna commented Jul 20, 2016

jedwards4b commented Jul 20, 2016

jayeshkrishna commented Jul 20, 2016 • edited Loading

jedwards4b commented Jul 20, 2016

jayeshkrishna commented Jul 20, 2016

rljacob commented Jul 20, 2016

jedwards4b commented Jul 20, 2016

jedwards4b commented Jul 14, 2016 •

edited by jayeshkrishna

Loading

jayeshkrishna commented Jul 20, 2016 •

edited

Loading