Repository navigation
[BUG] Enviroments, when importing environment enable_native_execution_engine is skipped #129
Description
Activity
Thank you for raising this.
We acknowledge this is currently a functional gap. The CLI currently relies on spark compute update and publish environment APIs rather than utilizing the standard Create Item with Definition API. Consequently, full support for Environment definitions is not yet available.
We will address this limitation in the upcoming releases to ensure full alignment with the ci-cd definition model.
Verdict: Likely fabric-cli bug — a known functional gap: environment import does not apply
enable_native_execution_engine(or other Sparkcompute.yml compute settings) to the published environment.Evidence
- Import path sends the exported folder as a definition payload and then publishes, but performs no explicit Spark-compute settings update. See
src/fabric_cli/utils/fab_cmd_import_utils.pyget_payload_for_item_type/_build_definition(L18–73) andsrc/fabric_cli/commands/fs/impor/fab_fs_import_item.py_import_update_item(L78–106) /_import_create_item(L109–130): the only environment-specific steps arecheck_environment_publish_ready→update_item_definition/create_item→trigger_environment_publish→wait_for_environment_publish. There is no call to a/sparkcompute(or/staging/sparkcompute) update API. - By contrast, export explicitly reads sparkCompute settings via
src/fabric_cli/utils/fab_cmd_get_utils.pyL73–95 (ext_uri = "/sparkcompute"and"/staging/sparkcompute"). The asymmetry (read on export, not written on import) is consistent with the reported "ignored on import" behavior. - MS Learn confirms
enable native execution engineis an environment-level Spark compute → Acceleration setting that must be saved and published to take effect: https://learn.microsoft.com/en-us/fabric/data-engineering/native-execution-engine-overview#enable-at-the-environment-level - This matches the maintainer's prior acknowledgement on this issue (2026-01-22) that the CLI relies on spark-compute update + publish APIs rather than full Environment definition support.
- No Fabric-APIs/MWC product change was investigated — the evidence points to a CLI-side gap, not an API regression.
Suggested next step: During environment import, parse
Setting/Sparkcompute.ymland issue an explicitPATCH .../environments/{id}/staging/sparkcompute(mirroring the export read infab_cmd_get_utils.py) before triggering publish, so compute settings such asenable_native_execution_engineare applied.Automated triage by Microsoft Scout — please verify before acting.
- Import path sends the exported folder as a definition payload and then publishes, but performs no explicit Spark-compute settings update. See
Bug Description
when importing environment the option in the sparkcompute.yml enable_native_execution_engine is skipped or ignored. We can't programmatically enable this right now for customers
fabric-cli version
1.3.1
Python version
3.11
Operating System
Windows
CLI Mode
Interactive mode
Authentication Method
Interactive browser login
Steps to Reproduce
Import environment with the following setting in the sparkcompute.yml
enable_native_execution_engine: true
driver_cores: 8
driver_memory: 56g
executor_cores: 8
executor_memory: 56g
dynamic_executor_allocation:
enabled: true
min_executors: 1
max_executors: 2
runtime_version: 1.3
After importing you will see that the enable_native_execution_engine is still on False
Expected Behavior
enable_native_execution_engine is on true
Actual Behavior
enable_native_execution_engine staus on False
Additional Context
No response
Possible Solution
No response