Add vdsm CI workflow with golden image caching

.github/workflows/vdsm.yml

@@ -0,0 +1,85 @@
+name: vdsm
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+env:
+  # Pins JavaScript-action Node runtime to Node 24 across all JS-based
+  # GitHub Actions in this workflow. Matches the pattern used by the
+  # other workflows in .github/workflows/ so CI behavior is consistent
+  # across runners during GitHub's staged Node runtime rollouts.
+  FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: true
+  DSM_VERSION: "7.2.2"
+  # Default admin credentials used inside vdsm setup. NOT secrets — the
+  # VM is ephemeral and exposed only to the test runner.
+  VDSM_ADMIN_USER: mcpadmin
+  VDSM_ADMIN_PASSWORD: "McpTest123!"
+
+jobs:
+  vdsm:
+    name: vdsm integration tests
+    runs-on: ubuntu-24.04
+    # Non-blocking while the job stabilizes. Promote to required check
+    # after the job is green on main for several consecutive runs.
+    continue-on-error: true
+    timeout-minutes: 30
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Verify /dev/kvm is available
+        run: |
+          ls -la /dev/kvm
+          sudo chmod 666 /dev/kvm
+          echo "KVM available"
+
+      - uses: astral-sh/setup-uv@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install project with vdsm extras
+        run: uv sync --extra dev --extra vdsm
+
+      - name: Install Playwright Chromium
+        run: uv run playwright install --with-deps chromium
+
+      # Cache key deliberately hashes only the DSM version and the Python
+      # files that drive image construction. `pyproject.toml` and `uv.lock`
+      # are NOT in the key because the image content is DSM-determined — a
+      # testcontainers or playwright dependency upgrade doesn't change the
+      # bits inside the saved tarball. If that assumption ever changes
+      # (e.g., a dep upgrade affects bind-mount format or timing), bump the
+      # cache key manually or add the manifest to hashFiles.
+      - name: Restore golden image cache
+        id: cache-golden
+        uses: actions/cache@v4
+        with:
+          path: |
+            .vdsm/golden/dsm-${{ env.DSM_VERSION }}.tar.gz
+            .vdsm/golden/dsm-${{ env.DSM_VERSION }}.meta.json
+          key: vdsm-golden-${{ env.DSM_VERSION }}-${{ hashFiles('scripts/vdsm_setup.py', 'tests/vdsm/setup_dsm.py', 'tests/vdsm/config.py', 'tests/vdsm/golden_image.py', 'tests/vdsm/container.py') }}
+
+      - name: Build golden image (cache miss)
+        if: steps.cache-golden.outputs.cache-hit != 'true'
+        run: |
+          uv run python scripts/vdsm_setup.py \
+            --version "${DSM_VERSION}" \
+            --admin-user "${VDSM_ADMIN_USER}" \
+            --admin-password "${VDSM_ADMIN_PASSWORD}" \
+            --yes
+
+      - name: Run vdsm integration tests
+        run: uv run pytest -m vdsm -v --no-cov --log-cli-level=INFO
+
+      - name: Upload logs on failure
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: vdsm-failure-logs
+          path: |
+            .vdsm/storage/**/log/
+            /tmp/vdsm-*.log
+          if-no-files-found: ignore
+          retention-days: 7

CHANGELOG.md

@@ -4,6 +4,11 @@
 
 ### Added
 
+- **vdsm CI workflow** (#24) — adds `.github/workflows/vdsm.yml` that runs the 47 vdsm integration tests on every PR using GitHub Actions' `ubuntu-24.04` runner with `/dev/kvm` access. Golden image is cached via `actions/cache@v4`, keyed on DSM version + hash of the setup scripts (`scripts/vdsm_setup.py`, `tests/vdsm/setup_dsm.py`, `tests/vdsm/config.py`, `tests/vdsm/golden_image.py`, `tests/vdsm/container.py`). Cache miss path invokes `scripts/vdsm_setup.py --yes` to build a fresh golden image. The new workflow is independent from `ci.yml` so a vdsm flake never blocks unit-test merges, and starts with `continue-on-error: true` until it has a track record of stability. Also adds a `--yes/-y` flag to `vdsm_setup.py` for non-interactive CI use.
+
+### Fixed
+
+- **Remove `synoshare --setopt` recycle bin enablement** (#24) — reverts the recycle bin enablement added in #23. DSM 7.2.2's `synoshare` CLI has no `--setopt` subcommand, so the call always fails (only hidden locally because the pre-existing golden image was built before the change). The revert is safe because `list_recycle_bin` in production code already handles a disabled recycle bin gracefully (returns a friendly "not enabled" message), which is exactly the path `test_02_list_recycle_bin` exercises. `TestRecycleBin` docstring updated to reflect that the test is tolerant of both states.
 - **vdsm 47/47: fix all 5 remaining virtual-dsm test failures** (#23) — fixes the 5 vdsm-specific test failures identified in the #22 handoff. Production code improvements: `get_dir_size` now handles DSM error 599 (task completed before status poll) gracefully instead of crashing, returning a best-effort result; `list_recycle_bin` catches error 408 on the `#recycle` path and returns a friendly "recycle bin not enabled" message instead of raising. Test fixes: `test_get_system_info` makes Temperature assertion conditional on non-virtual hardware; `test_search_keyword_finds_directory` creates a "Bambu Studio" directory via the API (DSM search matches names, not content) and searches from the share root with retries; `test_utilization_before_and_during_load` tolerates DirSize failure since it's only a load generator. Setup fix: `setup_dsm.py` enables recycle bin via `synoshare --setopt` after share creation. New unit test for the error 599 path.
 
 - **vdsm full automation: SSH + synoshare for shared folders** (#22) — completes the vdsm golden image setup by SSH-ing into the DSM guest VM (not the QEMU host container) to run `/usr/syno/sbin/synoshare --add` for proper DSM shared folder registration. Exposes SSH port 22, enables SSH via `SYNO.Core.Terminal` API, creates test data without sudo. Fixes FileStation API error 119 by adding `session` parameter to login. 42/47 vdsm tests pass; 5 remaining failures are virtual-dsm behavioral differences (no temp sensor, background task timing, search indexing, recycle bin config).

scripts/vdsm_setup.py

@@ -72,7 +72,15 @@ def _check_prerequisites() -> None:
     confirmation_prompt=True,
     help="Admin password for DSM setup wizard.",
 )
-def setup(dsm_version: str, admin_user: str, admin_password: str) -> None:
+@click.option(
+    "--yes",
+    "-y",
+    "assume_yes",
+    is_flag=True,
+    default=False,
+    help="Skip confirmation prompts (e.g. overwrite existing golden image). Required for CI use.",
+)
+def setup(dsm_version: str, admin_user: str, admin_password: str, assume_yes: bool) -> None:
     """Create a golden image for virtual-dsm testing.
 
     This command starts a virtual-dsm container, waits for you to complete
@@ -86,8 +94,9 @@ def setup(dsm_version: str, admin_user: str, admin_password: str) -> None:
     click.echo("  Docker: OK")
 
     # 2. Check if golden image already exists
-    if has_golden_image(dsm_version) and not click.confirm(
-        f"\nGolden image for DSM {dsm_version} already exists. Overwrite?"
+    if has_golden_image(dsm_version) and not (
+        assume_yes
+        or click.confirm(f"\nGolden image for DSM {dsm_version} already exists. Overwrite?")
     ):
         click.echo("Aborted.")
         sys.exit(0)

tests/test_integration.py

@@ -628,8 +628,12 @@ async def test_10_verify_deleted(self, nas_client: Any) -> None:
 class TestRecycleBin:
     """Test recycle bin listing after delete.
 
-    Requires the writable_folder's share to have recycle bin enabled.
-    Creates a folder, deletes it, checks recycle bin, then cleans up.
+    Works whether or not the writable_folder's share has recycle bin enabled:
+    when enabled, the deleted folder shows up in `#recycle`; when disabled,
+    `list_recycle_bin` returns a friendly "not enabled" message. Both paths
+    are valid and the test verifies `list_recycle_bin` returns a string in
+    either case. Creates a folder, deletes it, checks recycle bin, then
+    cleans up.
     """
 
     _RECYCLE_TEST = "_recycle_bin_test"

tests/vdsm/setup_dsm.py

@@ -485,15 +485,12 @@ def ssh(cmd: str, *, sudo: bool = True) -> tuple[int, str]:
             msg = f"synoshare --add {name} failed (rc={rc}): {out}"
             raise RuntimeError(msg)
 
-    # Enable recycle bin on shares (synoshare --add creates them disabled).
-    # This is a prerequisite for test_02_list_recycle_bin — failure is fatal.
-    for name in ["testshare", "writable"]:
-        rc, out = ssh(f"{_SYNOSHARE} --setopt {name} enable_recycle_bin=yes")
-        if rc == 0:
-            print(f"    Recycle bin enabled on {name}")
-        else:
-            msg = f"synoshare --setopt {name} enable_recycle_bin=yes failed (rc={rc}): {out}"
-            raise RuntimeError(msg)
+    # Note: DSM 7.2.2's `synoshare` CLI has no supported command to toggle the
+    # per-share recycle bin. Shares created via `synoshare --add` have recycle
+    # bin disabled. This is fine — `list_recycle_bin` in production code
+    # handles the disabled case gracefully (returns a friendly
+    # "Recycle bin is not enabled" message), and `test_02_list_recycle_bin`
+    # exercises exactly that path.
 
     # Verify shares registered
     rc, out = ssh(f"{_SYNOSHARE} --enum ALL")