Speed up AINode CI by consolidating tests and caching PyInstaller output

[JackieTien97]

Description

Speed up the Cluster IT - 1C1D1A CI pipeline (~52min → ~15-20min) via two optimizations:

1. Test consolidation (saves ~20min)

Merge 5 AINode IT test classes into a single AINodeSharedClusterIT that shares one 1C1D1A cluster:

• AINodeDeviceManageIT
• AINodeModelManageIT
• AINodeCallInferenceIT
• AINodeForecastIT
• AINodeInstanceManagementIT

This reduces cluster startups from 8 to 3 (SharedCluster + ClusterConfig + ConcurrentForecast).

Also converts AINodeClusterConfigIT from @Before/@After to @BeforeClass/@AfterClass, eliminating one redundant cluster restart by merging both dialect tests into a single method.

The 1C1D1A cluster startup takes ~3.5min each time (ConfigNode + DataNode + AINode), so eliminating 5 restarts saves ~17.5min. The actual test execution time is only ~9min out of the original 52min total.

2. PyInstaller dist caching (saves ~11min)

Added hash-based caching to build_binary.py:

• Computes SHA256 of all AINode source files, pyproject.toml, poetry.lock, and ainode.spec
• Caches the dist/ output at ~/.cache/iotdb-ainode-build/dist-cache/ (outside project dir, survives mvn clean)
• On cache hit, restores dist/ directly and skips the entire PyInstaller analysis + packaging phase

The PyInstaller phase scans thousands of hidden imports from torch/transformers/numpy and takes ~11min. When AINode source hasn't changed, this is entirely redundant.

Testing

All existing test logic is preserved — the tests are reorganized, not changed. AINodeConcurrentForecastIT remains separate (different data setup and LOAD/UNLOAD side effects).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 40.43%. Comparing base (3145e83) to head (0002ce3).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##             master   #17687   +/-   ##
=========================================
  Coverage     40.42%   40.43%           
  Complexity     2574     2574           
=========================================
  Files          5179     5179           
  Lines        349261   349262    +1     
  Branches      44683    44683           
=========================================
+ Hits         141206   141232   +26     
+ Misses       208055   208030   -25     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[JackieTien97]

Code Review

Verdict: LGTM

The test logic is faithfully preserved — this is a reorganization, not a behavioral change. The caching implementation is sound.

---

Observations

Test Consolidation (AINodeSharedClusterIT.java)

• All test methods from the original 5 classes are present and structurally identical.
• @BeforeClass correctly calls both prepareDataInTree() and prepareDataInTable() to satisfy all tests.
• AINodeConcurrentForecastIT correctly left separate (different data setup, LOAD/UNLOAD side effects).
• AINodeClusterConfigIT correctly left separate (it does REMOVE AINODE which would break the shared cluster).

AINodeClusterConfigIT refactoring

• Conversion from @Before/@After to @BeforeClass/@AfterClass eliminates one redundant cluster restart.
• The merged test now verifies both dialects in a single method with proper sequencing. Cleaner than the original.

PyInstaller caching (build_binary.py)

• compute_source_hash correctly excludes build/, dist/, __pycache__; includes all relevant sources + poetry.lock.
• shutil.copytree(..., symlinks=True) is the right call for PyInstaller output with symlinked shared libs.
• Hashes both relative path and file content — prevents collisions from file renames.

---

One Suggestion (non-blocking)

Consider adding Python version to the cache hash to avoid stale cache if the interpreter gets upgraded on CI:

hasher.update(sys.version.encode())

Since poetry.lock is already hashed, PyInstaller version changes are covered. But Python interpreter version is not reflected in any of the hashed files.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud