Harmon/insar-management-system-v2
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
3474e258399e03f2e09f292b6f136c13ca47b0cc
insar-management-system-v2/experiments/isce2_sbas_timeseries/notes/PHASE4_UNIFIED_ENV_EXPERIMENT.md
T

7.2 KiB
Raw Blame History

Phase 4 Unified-Environment Experiment

Updated: 2026-04-06

Goal

Validate whether the current WSL isce2 runtime can be cloned and extended with MintPy so that the SBAS experiment can run without the temporary isce bridge helper.

This is still an experiment-layer task.

Do not change the production backend yet.

Why run this phase

The bridge-based route is already validated, but a unified environment may be cleaner because:

  • many ISCE2 + MintPy users operate in one environment
  • command invocation becomes simpler
  • future worker deployment may be easier if one runtime is stable

The bridge-based route still remains the fallback baseline until this phase is verified.

Current Known Starting Point

WSL distro:

  • Ubuntu-24.04

Current environments observed on 2026-04-06:

  • isce2
  • mintpy

Observed package state:

  • isce2 env:
    • isce2 2.6.4
    • h5py 3.15.1
    • mintpy not installed
  • dedicated mintpy env:
    • mintpy 1.6.3

Initial Hypothesis

Expected best-case outcome:

  • clone isce2 into isce2_mintpy
  • install mintpy directly into the clone
  • reuse the same repo-local strict-mask and patched-launcher helpers
  • run the same LT-1 smoke test without the isce bridge wrapper

Main risk areas:

  • package solver may replace or downgrade key ISCE2-side numeric dependencies
  • MintPy may still require the same repo-local runtime workaround even in a unified env
  • GDAL / h5py / pyaps3 dependency changes may alter the known-good stack behavior

Reproducible Commands

1. Bootstrap the unified env

wsl -d Ubuntu-24.04 env TARGET_ENV=isce2_mintpy_v1 BOOTSTRAP_MODE=recreate USE_TUNA_MIRROR=1 MINTPY_SPEC=mintpy=1.6.3 bash /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scripts/install_mintpy_into_cloned_isce2_env_ubuntu2404.sh

Why BOOTSTRAP_MODE=recreate:

  • direct conda create --clone was not stable enough for this machine
  • it still followed source-package URLs and hit channel/TOS friction
  • the successful path was:
    • export the current isce2 dependency list
    • recreate the env through Tsinghua mirror channels
    • reinstall exported pip packages
    • install MintPy into the recreated env

Optional environment override:

TARGET_ENV=isce2_mintpy_v1 MINTPY_SPEC='mintpy=1.6.3'

2. Run MintPy commands directly inside the unified env

wsl -d Ubuntu-24.04 env MINTPY_ENV=isce2_mintpy_v1 bash /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scripts/run_mintpy_unified_env_ubuntu2404.sh prep_isce.py -h

3. Re-run the current LT-1 SBAS smoke test in the unified env

wsl -d Ubuntu-24.04 env MINTPY_ENV=isce2_mintpy_v1 bash /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scripts/run_mintpy_sbas_unified_env_smoketest_ubuntu2404.sh /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/configs/sample_smallbaseline_lt1_e123p3_n46p1.cfg /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/stack_work/mintpy_sbas_unified_v1

4. Export publish bundle in the unified env

wsl -d Ubuntu-24.04 env MINTPY_ENV=isce2_mintpy_v1 bash /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scripts/export_mintpy_publish_products_unified_env_ubuntu2404.sh /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/stack_work/mintpy_sbas_unified_v1 /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/publish/mintpy_sbas_unified_v1

Comparison Checklist

When this phase is executed, compare it against the bridge route on:

  • package versions after install
  • whether prep_isce.py imports cleanly without bridge
  • whether load_data succeeds
  • whether the strict-mask step is still required
  • whether the patched launcher is still required
  • whether output files match the existing bridge-based artifact set
  • whether geocode/export still succeeds from the unified env

Current Status

Successful environment:

  • /home/administrator/miniconda3/envs/isce2_mintpy_v1

Observed package/version state in the successful unified env:

  • conda list shows:
    • mintpy 1.6.3
  • runtime mintpy.__version__ reports:
    • 1.6.2
  • isce import path:
    • /home/administrator/miniconda3/envs/isce2_mintpy_v1/lib/python3.11/site-packages/isce/__init__.py
  • mintpy import path:
    • /home/administrator/miniconda3/envs/isce2_mintpy_v1/lib/python3.11/site-packages/mintpy/__init__.py
  • h5py:
    • 3.15.1

Successful unified-env SBAS work directory:

  • /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/stack_work/mintpy_sbas_unified_v1

Successful unified-env publish directory:

  • /mnt/z/Code/Insar_management_system_v2/experiments/isce2_sbas_timeseries/scratch/lt1a_strip1_hh_descending_e123p3_n46p1/publish/mintpy_sbas_unified_v1

Validated in the unified env:

  • prep_isce.py -h works without the isce bridge
  • load_data succeeded
  • strict-mask generation succeeded
  • modify_network -> velocity succeeded
  • publish export succeeded through geocoded HDF5, GeoTIFF, preview, and manifest.json

Observed unified-env output set under mintpy_sbas_unified_v1/:

  • timeseries.h5
  • velocity.h5
  • temporalCoherence.h5
  • maskTempCoh.h5
  • maskAllValid.h5
  • avgSpatialCoh.h5
  • numTriNonzeroIntAmbiguity.h5
  • numTriNonzeroIntAmbiguity.png

Observed publish bundle under publish/mintpy_sbas_unified_v1/:

  • manifest.json
  • assets/geo_timeseries.h5
  • assets/geo_velocity.h5
  • assets/geo_temporalCoherence.h5
  • assets/geo_maskTempCoh.h5
  • assets/velocity.tif
  • assets/temporalCoherence.tif
  • assets/maskTempCoh.tif
  • preview/velocity_preview.png
  • preview/numTriNonzeroIntAmbiguity.png
  • metadata/smallbaselineApp.cfg
  • metadata/source_quality_summary.json

Quality summary matched the bridge-based route:

  • maskAllValid:
    • 1219001 / 4076199
    • 29.91%
  • maskTempCoh:
    • 62987 / 4076199
    • 1.55%

Still required in the unified env:

  • strict maskAllValid.h5 before inversion
  • repo-local patched smallbaselineApp launcher

Prepared:

  • unified-env bootstrap script
  • unified-env MintPy runner
  • unified-env SBAS smoke-test runner
  • unified-env publish-export wrapper

Completed:

  • actual clone + install execution
  • actual smoke-test result capture
  • actual publish-export capture

Pending:

  • deeper comparison of output metadata against the bridge route
  • decide whether unified env or bridge env should be the default production candidate
  • decide whether to pin the runtime to the conda package label 1.6.3 or the internal MintPy version string 1.6.2

Current Judgment

At experiment level, the unified environment is now viable.

This phase confirms:

  • the current LT-1 SBAS route does not fundamentally require the isce bridge
  • a recreated isce2 + mintpy WSL env can complete:
    • MintPy load/inversion
    • geocode/export
    • publish-bundle generation

Current recommendation:

  • keep the bridge route as the already-known baseline until a fuller diff is written
  • but treat the unified env as a valid candidate for the future production runtime
Reference in New Issue View Git Blame Copy Permalink