The GitHub Actions job "Required Checks" on texera.git/feat/geesefs-helm-packaging has failed. Run started by GitHub user aicam (triggered by aicam).
Head commit for run: 4d99cf4bdf9920bc0ff816790c17766705b63040 / ali <[email protected]> feat(engine): mount a LakeFS repo per Python UDF via the JWT S3 proxy Wires the engine consumer that activates in-pod LakeFS-repository mounting, so a Python UDF can read a repository's files (a dataset or a model) from a local path. - PythonUDFOpDescV2: new "Mount dataset" property (/ownerEmail/datasetName/ versionName); at compile time it is resolved to a repository:commitHash locator and carried on the PhysicalOp. - FileResolver.resolveDatasetVersion: DB lookup mapping the path to (repositoryName, versionHash). - PhysicalOp / WorkerConfig: carry the locator to the Python worker. - DatasetMountManager: before the Python worker starts, exchange the pod's per-user JWT for a mount session at file-service and run GeeseFS against the JWT S3 proxy (no global LakeFS credentials in the pod); idempotent per (repository, commit) for the pod lifetime. - PythonWorkflowWorker: mount before spawning the worker and pass the path in its startup config; the Python side (texera_run_python_worker.py, executor_manager.py) exposes it to UDF code as MOUNTED_DATASET_PATH. End-to-end on minikube: a Python UDF loaded a ~2 GB sharded PyTorch model from the mount via torch.load() with bit-exact output, shards streaming as ranged reads through the proxy, and no LakeFS credentials in the pod. Part of #6606. Closes #6606. Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]> Claude-Session: https://claude.ai/code/session_014GjAdd2Q7og15uDrzLLUF8 Signed-off-by: ali <[email protected]> Report URL: https://github.com/apache/texera/actions/runs/29781744422 With regards, GitHub Actions via GitBox
