Testing zppy for E3SM Unified release

This tutorial covers testing zppy for an E3SM Unified release. Examples are from testing for E3SM Unified v1.5.0.

Parameters

Only a few parameters in the configuration file need to change for different machines. Be sure to change the file paths to fit your own workspaces.

Warning

Also consider the space quotas or purge policies for the file paths you use. You’ll be storing about 2 TB of data.

Chrysalis:

  • <INPUT>: /lcrc/group/e3sm/ac.forsyth2/E3SM_simulations/e3sm_unified_test_simulation/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis

  • <OUTPUT>: /lcrc/group/e3sm/ac.forsyth2/E3SM_simulations/e3sm_unified_test_simulation/

  • <WWW>: /lcrc/group/e3sm/public_html/diagnostic_output/ac.forsyth2/E3SM/e3sm_unified_test_simulation/

  • <PARTITION>: compute

  • <ENVIRONMENT_COMMANDS>: source /lcrc/soft/climate/e3sm-unified/test_e3sm_unified_1.5.0rc8_chrysalis.sh

  • <MAPPING_FILE>: /home/ac.zender/data/maps/map_ne30pg2_to_cmip6_180x360_aave.20200201.nc

  • <REFERENCE_DATA_PATH>: /lcrc/soft/climate/e3sm_diags_data/obs_for_e3sm_diags/climatology

  • <OBS_TS>: /lcrc/soft/climate/e3sm_diags_data/obs_for_e3sm_diags/time-series

  • <DC_OBS_CLIMO>: /lcrc/group/e3sm/public_html/e3sm_diags_test_data/unit_test_complete_run/obs/climatology

Compy:

  • <INPUT>: /compyfs/fors729/e3sm_unified_test_zppy/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis

  • <OUTPUT>: /compyfs/fors729/e3sm_unified_test_zppy/

  • <WWW>: /compyfs/www/fors729/E3SM/e3sm_unified_test_simulation/

  • <PARTITION>: slurm

  • <ENVIRONMENT_COMMANDS>: source /share/apps/E3SM/conda_envs/test_e3sm_unified_1.5.0rc8_compy.sh

  • <MAPPING_FILE>: /compyfs/zender/maps/map_ne30pg2_to_cmip6_180x360_aave.20200201.nc

  • <REFERENCE_DATA_PATH>: /compyfs/e3sm_diags_data/obs_for_e3sm_diags/climatology

  • <OBS_TS>: /compyfs/e3sm_diags_data/obs_for_e3sm_diags/time-series

  • <DC_OBS_CLIMO>: /compyfs/e3sm_diags_data/obs_for_e3sm_diags/climatology

Cori:

  • <INPUT>: /global/cscratch1/sd/forsyth/e3sm_unified_test_zppy/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis

  • <OUTPUT>: /global/cscratch1/sd/forsyth/e3sm_unified_test_zppy/

  • <WWW>: /global/cfs/cdirs/e3sm/www/forsyth/E3SM/e3sm_unified_test_simulation/

  • <PARTITION>: haswell

  • <ENVIRONMENT_COMMANDS>: source /global/common/software/e3sm/anaconda_envs/test_e3sm_unified_1.5.0rc8_cori-haswell.sh

  • <MAPPING_FILE>: /global/homes/z/zender/data/maps/map_ne30pg2_to_cmip6_180x360_aave.20200201.nc

  • <REFERENCE_DATA_PATH>: /global/cfs/cdirs/e3sm/acme_diags/obs_for_e3sm_diags/climatology

  • <OBS_TS>: /global/cfs/cdirs/e3sm/acme_diags/obs_for_e3sm_diags/time-series

  • <DC_OBS_CLIMO>: /global/cfs/cdirs/e3sm/acme_diags/obs_for_e3sm_diags/climatology

These values don’t appear in the configuration file but are still useful to specify here:

Chrysalis:

Compy:

Cori:

Testing zppy on a small simulation

1. Find a simulation to test on. In this example, we’ll use a simulation on Chrysalis: /lcrc/group/e3sm/ac.golaz/E3SM_simulations/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis/.

2. The simulation might be too large to reasonably test on. We can copy over a subsection to a testing directory with a script like the following. In this example, the simulation data starts at year 51. We want to have at least 20 years, so we’ll copy over years 51-79 using the pattern *.00[5-7]*. This subsection turns out to be about 2 terabytes in size.

#!/bin/bash

input=/lcrc/group/e3sm/ac.golaz/E3SM_simulations/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis/
output=e3sm_unified_test_simulation/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis/

# For E3SM Diags: About 40 minutes to run
# Copy over years xx5x - xx7x
cp ${input}archive/atm/hist/*.00[5-7]* ${output}archive/atm/hist
cp ${input}archive/cpl/hist/*.00[5-7]* ${output}archive/cpl/hist
cp ${input}archive/ice/hist/*.00[5-7]* ${output}archive/ice/hist
cp ${input}archive/lnd/hist/*.00[5-7]* ${output}archive/lnd/hist
cp ${input}archive/ocn/hist/*.00[5-7]* ${output}archive/ocn/hist
cp ${input}archive/rof/hist/*.00[5-7]* ${output}archive/rof/hist

# For MPAS-Analysis: Less than a minute to run
cp ${input}run/mpaso_in ${output}run/mpaso_in
cp ${input}run/mpassi_in ${output}run/mpassi_in
cp ${input}run/streams.ocean ${output}run/streams.ocean
cp ${input}run/streams.seaice ${output}run/streams.seaice
cp ${input}run/*rst* ${output}run/

3. Now we need to load zppy. We first want to test out zppy from the new unified environment release candidate. We can do this by running <ENVIRONMENT_COMMANDS>. See 3a-c below if errors are encountered in later steps.

3a. If zppy has problems in later steps, we’ll have to make changes in the zppy code and create a new release candidate; until the new zppy release candidate is included in the new unified environment release candidate, we’ll have to just use the zppy development environment. This will require running pip install . in that environment from the top level of the repo.

3b. It may be the case that the configuration file had an error and thus neither zppy nor any package that zppy runs actually failed. In that case, fix the configuration file.

3c. Keep in mind if you run zppy -c multiple times, it will only rerun tasks that have failed. This may or may not be what we want. For example, if we change the environment we’re using, we probably want to rerun everything using the new environment.

  1. Create a configuration file e3sm_unified_test_simulation.cfg.

 1[default]
 2input = <INPUT>
 3input_subdir = archive/atm/hist
 4# The directory that the `post` directory is placed in.
 5output = <OUTPUT>
 6case = 20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis
 7www = <WWW>
 8e3sm_unified = latest
 9partition = <PARTITION>
10environment_commands = "<ENVIRONMENT_COMMANDS>"
11
12[climo]
13active = True
14years = "51:70:20",
15# Use default vars
16
17  [[ atm_monthly_180x360_aave ]]
18  mapping_file = <MAPPING_FILE>
19  frequency = "monthly"
20
21  [[ atm_monthly_diurnal_8xdaily_180x360_aave ]]
22  input_subdir = "archive/atm/hist"
23  input_files = "eam.h4"
24  mapping_file = <MAPPING_FILE>
25  vars = "PRECT"
26  frequency = "monthly_diurnal_8xdaily"
27
28[ts]
29active = True
30years = "51:70:10",
31
32  [[ atm_monthly_180x360_aave ]]
33  input_subdir = "archive/atm/hist"
34  input_files = "eam.h0"
35  frequency = "monthly"
36  mapping_file = <MAPPING_FILE>
37  # Use default vars
38
39  [[ atm_daily_180x360_aave ]]
40  input_subdir = "archive/atm/hist"
41  input_files = "eam.h1"
42  frequency = "daily"
43  mapping_file = <MAPPING_FILE>
44  vars = "PRECT"
45
46  [[ atm_monthly_glb ]]
47  input_subdir = "archive/atm/hist"
48  input_files = "eam.h0"
49  frequency = "monthly"
50  mapping_file = "glb"
51  # Use default vars
52
53  [[ land_monthly ]]
54  input_subdir = "archive/lnd/hist"
55  input_files = "elm.h0"
56  frequency = "monthly"
57  mapping_file = <MAPPING_FILE>
58  vars = "FSH,RH2M"
59
60[e3sm_diags]
61active = True
62years = "51:70:20",
63ts_num_years = 10
64ref_start_yr = 1979
65ref_final_yr = 2016
66
67  [[ atm_monthly_180x360_aave ]]
68  short_name = "20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis"
69  grid = "180x360_aave"
70  reference_data_path = "<REFERENCE_DATA_PATH>"
71  obs_ts = "<OBS_TS>"
72  dc_obs_climo = "<DC_OBS_CLIMO>"
73  climo_diurnal_subsection = "atm_monthly_diurnal_8xdaily_180x360_aave"
74
75[e3sm_diags_vs_model]
76active = False
77
78[amwg]
79active = False
80
81[mpas_analysis]
82active = True
83walltime = "24:00:00"
84parallelTaskCount = 6
85ts_years = "51-70",
86enso_years = "51-70",
87climo_years ="51-70",
88mesh = "EC30to60E2r2"
89anomalyRefYear = 51
90
91[global_time_series]
92active = True
93years = "51-70",
94ts_num_years = 10
95figstr=coupled_v2rc3e
96moc_file=mocTimeSeries_0051-0070.nc
97experiment_name=20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis
98ts_years = "51-70",
99climo_years ="51-70",

5. We can now test out zppy by running zppy -c e3sm_unified_test_simulation.cfg. If that command fails, return to step 3.

  1. If the above command completes successfully, run:
    # cd into output directory
cd <OUTPUT>/post/scripts
# Check status of files that either failed or are still running.
grep -v "OK" *status
# If there is an error, return to step 3.
# If you want to remove all output run the following two steps:
rm -rf <OUTPUT>/post
rm -rf <WWW>/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis

7. Output can be viewed at the web link corresponding to <WWW> in the configuration file (i.e., <HTML> defined in “Parameters” above).

Archive simulation using zstash

  1. Create a batch script zstash_create.bash to run zstash.
    #!/bin/bash

#SBATCH  --job-name=zstash_create
#SBATCH  --nodes=1
#SBATCH  --output=<ZSTASH_OUTPUT_DIR>/zstash_create.o%j
#SBATCH  --exclusive
#SBATCH  --time=04:00:00

<ENVIRONMENT_COMMANDS>
zstash create --hpss=none --cache=<ZSTASH_CACHE_DIR> <OUTPUT>/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis

  2. Run with sbatch zstash_create.bash. This example takes about 2.5 hours to run.

Transfer simulation to other machines

We can follow an adapted version of the steps at https://e3sm-project.github.io/zstash/_build/html/master/best_practices.html#transfer-to-nersc-hpss to transfer the data.

Login to Globus: https://www.globus.org/ (using NERSC credentials)

Transfer to Cori

  1. On the left sidebar, choose “ENDPOINTS”.

  2. Search for “NERSC DTN”. Click on Green power button to activate endpoint.

  3. On the left sidebar, choose “ENDPOINTS”.

4. Search for “lcrc#dtn_bebop”. Click on Green power button to activate endpoint. Log in using LCRC credentials.

5. Paste the path to the zstash archive (<ZSTASH_CACHE_DIR> for Chrysalis) in the “Path” box.

  1. Click “Transfer or Sync to…” on the right side. The screen will now be split.

  2. On the left side, choose “Select all”

  3. On the right side, put “NERSC DTN” for collection.

9. Paste the NERSC path you want the archive copied to (<ZSTASH_CACHE_DIR> for Cori). This path needs to already exist.

  1. Click “Transfer & Sync Options” in the center.

11. Choose “sync - only transfer new or changed files” (choose “modification time is newer” in the dropdown box), “preserve source file modification times”, and “verify file integrity after transfer”.

  1. For “Label This Transfer”, put something like “zstash archive LCRC to NERSC”.

  2. On the left side, click “Start >”. This will start the transfer from LCRC to NERSC.

Transfer to Compy

  1. On the left sidebar, choose “ENDPOINTS”.

2. Search for “pic#compy-dtn”. Click on Green power button to activate endpoint. Log in using Compy credentials.

  1. On the left sidebar, choose “File Manager.” The screen will now be split.

  2. On the left side, put “lcrc#dtn_bebop” for “Collection”.

5. On the left side, paste the path to the zstash archive (<ZSTASH_CACHE_DIR> for Chrysalis) in the “Path” box.

  1. On the left side, choose “Select all”.

  2. On the right side, put “pic#compy-dtn” for “Collection”.

8. Paste the Compy path you want the archive copied to (<ZSTASH_CACHE_DIR> for Compy). This path needs to already exist.

  1. Click “Transfer & Sync Options” in the center.

10. Choose “sync - only transfer new or changed files” (choose “modification time is newer” in the dropdown box), “preserve source file modification times”, and “verify file integrity after transfer”.

  1. For “Label This Transfer”, put something like “zstash archive LCRC to Compy”.

  2. On the left side, click “Start >”. This will start the transfer from LCRC to Compy.

Check transfers were successful

Cori

  1. Create a batch script zstash_check.bash:

    #!/bin/bash
#SBATCH  --job-name=zstash_check
#SBATCH  --nodes=1
#SBATCH  --output=<ZSTASH_OUTPUT_DIR>/zstash_check.o%j
#SBATCH  --exclusive
#SBATCH  --time=3:00:00
#SBATCH  -q regular
#SBATCH  --constraint=haswell

<ENVIRONMENT_COMMANDS>
zstash check --hpss=none --cache=<ZSTASH_CACHE_DIR>

  2. Run sbatch zstash_check.bash. This takes about an hour to run.

Compy

  1. Create a batch script zstash_check.bash:

    #!/bin/bash
#SBATCH  --job-name=zstash_check
#SBATCH  --nodes=1
#SBATCH  --output=<ZSTASH_OUTPUT_DIR>/zstash_check.o%j
#SBATCH  --exclusive
#SBATCH  --time=3:00:00

<ENVIRONMENT_COMMANDS>
zstash check --hpss=none --cache=<ZSTASH_CACHE_DIR>

  2. Run sbatch zstash_check.bash. This takes over 3 hours to run.

Extract the data from the archives

Use zstash extract.

Cori

  1. mkdir unified_test_extraction.

  2. Create a batch script zstash_extract.bash:

    #!/bin/bash
#SBATCH  --job-name=zstash_extract
#SBATCH  --nodes=1
#SBATCH  --output=<ZSTASH_OUTPUT_DIR>/zstash_extract.o%j
#SBATCH  --exclusive
#SBATCH  --time=4:00:00
#SBATCH  -q regular
#SBATCH  --constraint=haswell

<ENVIRONMENT_COMMANDS>
cd unified_test_extraction
zstash extract --hpss=none --cache=<ZSTASH_CACHE_DIR>

  3. Run sbatch zstash_extract.bash.

Compy

  1. mkdir unified_test_extraction.

  2. Create a batch script zstash_extract.bash:

    #!/bin/bash
#SBATCH  --job-name=zstash_extract
#SBATCH  --nodes=1
#SBATCH  --output=<ZSTASH_OUTPUT_DIR>/zstash_extract.o%j
#SBATCH  --exclusive
#SBATCH  --time=6:00:00

<ENVIRONMENT_COMMANDS>
cd unified_test_extraction
zstash extract --hpss=none --cache=<ZSTASH_CACHE_DIR>

  3. Run sbatch zstash_extract.bash.

Testing zppy on other machines

mv e3sm_unified_test_zstash/unified_test_extraction e3sm_unified_test_zppy/20210528.v2rc3e.piControl.ne30pg2_EC30to60E2r2.chrysalis.

Go through steps 3-7 of “Testing zppy on a small simulation” but for Cori (haswell) and Compy.

Sample configuration files for Compy and Cori

<OUTPUT>/e3sm_unified_test_simulation.cfg: Use the configuration file from “Testing zppy on a small simulation” step 4, updating the appropriate parameters for Compy or Cori.

Run unit tests and integration tests

Run the unit tests and integration tests on Chrysalis:

cd into your clone of the zppy repo. Then:

git checkout main # if not already on the `main` branch
git fetch upstream
git rebase upstream/main
<ENVIRONMENT_COMMANDS> # Load the appropriate environment
python -m unittest tests/test_*.py # Run unit tests
python -m unittest tests/integration/test_*.py # Run integration tests