fix: clean interrupted staging temporaries

deploy/production/lithium/compose.yaml

@@ -2,7 +2,7 @@ name: archive-control-archive
 
 services:
   archive-client:
-    image: sodium/archive-clients:v0.1.5
+    image: sodium/archive-clients:v0.1.8
     user: "1000:1000"
     restart: unless-stopped
     command: ["--config", "/etc/archive-control/client.toml"]

deploy/production/x1/client.toml

@@ -39,4 +39,7 @@ endpoint = "http://127.0.0.1:8384"
 api_key_file = "/run/secrets/syncthing_api_key"
 api_root = "/var/syncthing"
 local_root = "/data/sync"
+# This existing folder is physically inside the qB data tree. Map it through
+# that same client bind mount so source staging can use hardlinks.
+local_path_overrides = { "/var/syncthing/DownloadsSync" = "/data/qb/Sync" }
 advertised_addresses = ["dynamic"]

deploy/production/x1/compose.yaml

@@ -2,7 +2,7 @@ name: archive-control-cache
 
 services:
   archive-client:
-    image: sodium/archive-clients:v0.1.5
+    image: sodium/archive-clients:v0.1.8
     user: "1001:1001"
     restart: unless-stopped
     network_mode: host
@@ -16,4 +16,3 @@ services:
       - ./backups:/var/backups/archive-control
       - /home/ubuntu/Downloads:/data/qb
       - /home/ubuntu/compose/syncthing/st_home:/data/sync
-      - /home/ubuntu/Downloads/Sync:/data/sync/DownloadsSync

deploy/production/x2/compose.yaml

@@ -2,7 +2,7 @@ name: archive-control-cache
 
 services:
   archive-client:
-    image: sodium/archive-clients:v0.1.5
+    image: sodium/archive-clients:v0.1.8
     user: "1001:1001"
     restart: unless-stopped
     network_mode: host

docs/deployment-and-usage.md

@@ -125,6 +125,11 @@ local_root = "/data/sync"
 advertised_addresses = ["dynamic"]
 ```
 
+When an existing Syncthing folder is physically nested in the qB data root,
+use a `local_path_overrides` entry to map that exact Syncthing API path through
+the same client bind mount. This enables hardlinks without creating two Docker
+mount boundaries for the same host files.
+
 The remaining node examples omit optional `[connection]`, `[jobs]`, and
 `[backup]` tables and therefore use these same defaults; deployments may
 override them per node.
@@ -213,7 +218,7 @@ cache/archive routes according to policy.
 ```yaml
 services:
   archive-client:
-    image: sodium/archive-clients:v0.1.5
+    image: sodium/archive-clients:v0.1.8
     user: "1001:1001"
     restart: unless-stopped
     command: ["archive-client", "--config", "/etc/archive-control/client.toml"]

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "archive-clients"
-version = "0.1.5"
+version = "0.1.8"
 requires-python = ">=3.11"
 dependencies = ["protobuf==7.35.1", "websockets==16.0"]
 

src/archive_clients/config.py

@@ -27,17 +27,28 @@ _ENV = re.compile(r"\$\{([A-Za-z_][A-Za-z0-9_]*)\}")
 class RootMapping:
     api_root: PurePosixPath
     local_root: Path
+    local_path_overrides: tuple[tuple[PurePosixPath, Path], ...] = ()
 
     def api_to_local(self, api_path: str) -> Path:
         candidate = PurePosixPath(api_path)
-        try:
-            relative = candidate.relative_to(self.api_root)
-        except ValueError as exc:
-            raise ConfigError("API path is outside its configured root") from exc
-        if any(part in {"", ".", ".."} for part in relative.parts):
-            raise ConfigError("API path contains an unsafe component")
+        for api_root, local_root in self.local_path_overrides:
+            relative = _safe_relative(candidate, api_root)
+            if relative is not None:
+                return local_root.joinpath(*relative.parts)
+        relative = _safe_relative(candidate, self.api_root)
+        if relative is None:
+            raise ConfigError("API path is outside its configured root")
         return self.local_root.joinpath(*relative.parts)
 
+    def local_root_for_api(self, api_path: str) -> Path:
+        candidate = PurePosixPath(api_path)
+        for api_root, local_root in self.local_path_overrides:
+            if _safe_relative(candidate, api_root) is not None:
+                return local_root
+        if _safe_relative(candidate, self.api_root) is None:
+            raise ConfigError("API path is outside its configured root")
+        return self.local_root
+
 
 @dataclass(frozen=True)
 class ServiceConfig:
@@ -48,10 +59,13 @@ class ServiceConfig:
     password_file: Path | None = None
     api_key_file: Path | None = None
     advertised_addresses: tuple[str, ...] = ()
+    local_path_overrides: tuple[tuple[PurePosixPath, Path], ...] = ()
 
     @property
     def roots(self) -> RootMapping:
-        return RootMapping(self.api_root, self.local_root)
+        return RootMapping(
+            self.api_root, self.local_root, self.local_path_overrides
+        )
 
     def read_password(self) -> str | None:
         return (
@@ -160,7 +174,7 @@ def _service(value: Any, name: str) -> ServiceConfig:
         raise ConfigError(f"{name} must be a table")
     allowed = {
         "endpoint", "api_root", "local_root", "username", "password_file",
-        "api_key_file", "advertised_addresses",
+        "api_key_file", "advertised_addresses", "local_path_overrides",
     }
     _keys(value, allowed, name)
     api_root = PurePosixPath(_string(value, "api_root"))
@@ -182,6 +196,7 @@ def _service(value: Any, name: str) -> ServiceConfig:
         raise ConfigError("qbittorrent username and password_file are required")
     if name == "syncthing" and "api_key_file" not in value:
         raise ConfigError("syncthing api_key_file is required")
+    overrides = _local_path_overrides(value, api_root, name)
     return ServiceConfig(
         _endpoint(value, "endpoint", {"http", "https"}), api_root,
         _absolute_path(value, "local_root"), username,
@@ -189,10 +204,46 @@ def _service(value: Any, name: str) -> ServiceConfig:
         if "password_file" in value else None,
         _absolute_path(value, "api_key_file")
         if "api_key_file" in value else None,
-        tuple(addresses),
+        tuple(addresses), overrides,
     )
 
 
+def _local_path_overrides(
+    value: dict[str, Any], api_root: PurePosixPath, name: str
+) -> tuple[tuple[PurePosixPath, Path], ...]:
+    raw = value.get("local_path_overrides", {})
+    if name != "syncthing" and raw:
+        raise ConfigError(f"{name}.local_path_overrides is unsupported")
+    if not isinstance(raw, dict):
+        raise ConfigError(f"{name}.local_path_overrides must be a table")
+    parsed: list[tuple[PurePosixPath, Path]] = []
+    for raw_api_path, raw_local_path in raw.items():
+        if not isinstance(raw_api_path, str) or not isinstance(raw_local_path, str):
+            raise ConfigError(f"{name}.local_path_overrides entries must be strings")
+        candidate = PurePosixPath(raw_api_path)
+        if not candidate.is_absolute() or ".." in candidate.parts:
+            raise ConfigError(f"{name}.local_path_overrides API path is invalid")
+        if _safe_relative(candidate, api_root) is None:
+            raise ConfigError(f"{name}.local_path_overrides API path is outside root")
+        local = Path(raw_local_path)
+        if not local.is_absolute():
+            raise ConfigError(f"{name}.local_path_overrides local path is invalid")
+        parsed.append((candidate, local))
+    return tuple(sorted(parsed, key=lambda item: len(item[0].parts), reverse=True))
+
+
+def _safe_relative(
+    candidate: PurePosixPath, root: PurePosixPath
+) -> PurePosixPath | None:
+    try:
+        relative = candidate.relative_to(root)
+    except ValueError:
+        return None
+    if any(part in {"", ".", ".."} for part in relative.parts):
+        raise ConfigError("API path contains an unsafe component")
+    return relative
+
+
 def _connection(value: Any) -> ConnectionConfig:
     if not isinstance(value, dict):
         raise ConfigError("connection must be a table")

src/archive_clients/jobs.py

@@ -4,6 +4,8 @@ from __future__ import annotations
 
 import hashlib
 import json
+import logging
+import os
 import shutil
 import stat
 import threading
@@ -26,10 +28,11 @@ from archive_clients.qbittorrent import (
 )
 from archive_clients.resources import NormalizedResource
 from archive_clients.state import ClientStore
-from archive_clients.syncthing import SyncthingTransferObserver
+from archive_clients.syncthing import RouteSetupError, SyncthingTransferObserver
 from archive_clients.transfer import (
     TransferError,
     TransferIntegrityError,
+    cleanup_partial_transfer,
     cleanup_transfer,
     load_published_transfer,
     materialize_transfer,
@@ -52,6 +55,9 @@ class JobCancelled(JobExecutionError):
     pass
 
 
+logger = logging.getLogger(__name__)
+
+
 class ClientJobExecutor:
     def __init__(
         self,
@@ -578,7 +584,17 @@ class ClientJobExecutor:
                 f"staged {completed} of {total} bytes",
             ),
         )
+        try:
             self._observer(definition).rescan()
+        except RouteSetupError as exc:
+            # A manual Syncthing scan may take longer than the bounded REST
+            # request timeout for a large newly linked file. The folder watch
+            # and the following transfer observation still converge, so do
+            # not roll back an otherwise durable staged transfer.
+            logger.warning("syncthing_rescan_deferred", extra={
+                "job_id": definition.job_id,
+                "error_type": type(exc).__name__,
+            })
 
     def _wait_for_syncthing(
         self,
@@ -818,6 +834,11 @@ class ClientJobExecutor:
             qb_root=job_directory,
             store=self.store,
         )
+        cleanup_partial_transfer(
+            job_directory,
+            job_id=definition.job_id,
+            store=self.store,
+        )
         for name in ("ready.json", "manifest.json"):
             candidate = job_directory / name
             if candidate.is_file() and not candidate.is_symlink():
@@ -893,6 +914,7 @@ class ClientJobExecutor:
             if (
                 not stat.S_ISREG(metadata.st_mode)
                 or metadata.st_dev != destination_device
+                or _mount_id(source) != _mount_id(destination_root)
             ):
                 required += logical_bytes
         return required
@@ -1143,6 +1165,8 @@ def _job_error_code(error: Exception) -> int:
         return common_pb2.ERROR_CODE_INTEGRITY_CHECK_FAILED
     if isinstance(error, PermissionError):
         return common_pb2.ERROR_CODE_PERMISSION_DENIED
+    if isinstance(error, RouteSetupError):
+        return common_pb2.ERROR_CODE_UNAVAILABLE
     if isinstance(error, JobExecutionError):
         return common_pb2.ERROR_CODE_PRECONDITION_FAILED
     if isinstance(error, TransferError):
@@ -1150,3 +1174,41 @@ def _job_error_code(error: Exception) -> int:
     if isinstance(error, EvictionError):
         return common_pb2.ERROR_CODE_PRECONDITION_FAILED
     return common_pb2.ERROR_CODE_INTERNAL
+
+
+def _mount_id(path: Path) -> str | None:
+    """Return Linux's effective mount ID for a path when procfs is available.
+
+    Bind mounts can share ``st_dev`` while still rejecting ``link(2)`` with
+    ``EXDEV``.  Mount IDs distinguish that case without creating probe files
+    inside a Syncthing folder.
+    """
+
+    try:
+        target = os.path.realpath(path)
+        best: tuple[int, str] | None = None
+        with open("/proc/self/mountinfo", encoding="utf-8") as source:
+            for line in source:
+                fields = line.rstrip("\n").split(" ")
+                if len(fields) < 5:
+                    continue
+                mountpoint = _unescape_mount_path(fields[4])
+                if target != mountpoint and not target.startswith(
+                    mountpoint.rstrip("/") + "/"
+                ):
+                    continue
+                candidate = (len(mountpoint), fields[0])
+                if best is None or candidate[0] > best[0]:
+                    best = candidate
+        return None if best is None else best[1]
+    except OSError:
+        return None
+
+
+def _unescape_mount_path(value: str) -> str:
+    return (
+        value.replace("\\040", " ")
+        .replace("\\011", "\t")
+        .replace("\\012", "\n")
+        .replace("\\134", "\\")
+    )

src/archive_clients/routes.py

@@ -37,7 +37,9 @@ def discover_routes(
             relative = normalized_api_path.relative_to(roots.api_root)
         except (ConfigError, ValueError):
             continue
-        if not _lexically_within(local_path, roots.local_root):
+        if not _lexically_within(
+            local_path, roots.local_root_for_api(normalized_api_path.as_posix())
+        ):
             continue
         if relative == PurePosixPath("."):
             continue

src/archive_clients/syncthing.py

@@ -356,7 +356,9 @@ class SyncthingRouteManager:
             local_path = self.config.roots.api_to_local(api_path)
         except ConfigError as exc:
             raise RoutePathConflict("route path is outside the sync root") from exc
-        resolved_root = self.config.local_root.resolve(strict=False)
+        resolved_root = self.config.roots.local_root_for_api(api_path).resolve(
+            strict=False
+        )
         try:
             local_path.resolve(strict=False).relative_to(resolved_root)
         except ValueError as exc:

src/archive_clients/transfer.py

@@ -456,6 +456,36 @@ def cleanup_transfer(job_directory: Path) -> bool:
     return True
 
 
+def cleanup_partial_transfer(
+    job_directory: Path,
+    *,
+    job_id: str,
+    store: ClientStore,
+) -> None:
+    """Remove copy/reflink temporaries left before a transfer is published.
+
+    A source-stage failure before ``manifest.json``/``ready.json`` exists
+    cannot use :func:`cleanup_transfer`. The file-operation journal is the
+    authoritative list of owned destinations, so derive each temporary path
+    from it rather than recursively removing arbitrary content from a shared
+    Syncthing folder.
+    """
+
+    for row in store.file_operation_rows(job_id):
+        intent = json.loads(str(row["intent_json"]))
+        destination = job_directory / _relative_path(str(intent["destination"]))
+        temporary = _temporary_path(destination, str(row["operation_id"]))
+        try:
+            metadata = temporary.lstat()
+        except FileNotFoundError:
+            continue
+        if not stat.S_ISREG(metadata.st_mode):
+            raise TransferIntegrityError(
+                "job-owned temporary cleanup path is not a regular file"
+            )
+        temporary.unlink()
+
+
 def canonical_message_json(message: object) -> bytes:
     value = json_format.MessageToDict(
         message,

tests/test_config.py

@@ -1,7 +1,7 @@
 import os
 import tempfile
 import unittest
-from pathlib import Path
+from pathlib import Path, PurePosixPath
 from unittest.mock import patch
 
 from archive_clients.config import ClientConfig, ConfigError, RootMapping
@@ -63,6 +63,21 @@ class ConfigTests(unittest.TestCase):
         with self.assertRaises(ConfigError):
             mapping.api_to_local("/elsewhere/file")
 
+    def test_mapping_can_override_one_syncthing_folder_locally(self):
+        mapping = RootMapping(
+            PurePosixPath("/sync"),
+            Path("/local/sync"),
+            ((PurePosixPath("/sync/DownloadsSync"), Path("/local/qb/Sync")),),
+        )
+        self.assertEqual(
+            mapping.api_to_local("/sync/DownloadsSync/job/ready.json"),
+            Path("/local/qb/Sync/job/ready.json"),
+        )
+        self.assertEqual(
+            mapping.local_root_for_api("/sync/DownloadsSync"),
+            Path("/local/qb/Sync"),
+        )
+
     def test_endpoint_scheme_and_job_keys_are_strict(self):
         with tempfile.TemporaryDirectory() as directory:
             root = Path(directory)

tests/test_jobs.py

@@ -11,6 +11,7 @@ from archive_clients.jobs import (
     JobExecutionError,
     _resource_fingerprint,
 )
+from archive_clients.syncthing import RouteSetupError
 from archive_clients.resources import normalize_resource
 from archive_clients.state import ClientStore
 from archive_control.v1 import control_pb2, job_pb2
@@ -31,6 +32,11 @@ class CompleteSyncthing:
         self.posts.append(path)
 
 
+class SlowRescanSyncthing(CompleteSyncthing):
+    def post(self, path):
+        raise RouteSetupError("Syncthing API is unavailable")
+
+
 class ClientJobHappyPathTests(unittest.TestCase):
     def test_capacity_guard_fails_before_data_movement(self):
         with tempfile.TemporaryDirectory() as directory:
@@ -78,6 +84,33 @@ class ClientJobHappyPathTests(unittest.TestCase):
                 content=b"x" * 4096,
             )
 
+    def test_mount_boundary_requires_copy_space_even_with_same_device(self):
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            source_root = root / "source"
+            destination_root = root / "destination"
+            source_root.mkdir()
+            destination_root.mkdir()
+            (source_root / "fixture.bin").write_bytes(b"fixture")
+            with patch(
+                "archive_clients.jobs._mount_id",
+                side_effect=("source-mount", "destination-mount"),
+            ):
+                required = ClientJobExecutor._copy_required_bytes(
+                    source_root,
+                    destination_root,
+                    (("fixture.bin", 7),),
+                )
+            self.assertEqual(required, 7)
+
+    def test_source_stage_survives_a_timed_out_syncthing_rescan(self):
+        with tempfile.TemporaryDirectory() as directory:
+            self._run_transfer(
+                Path(directory),
+                job_pb2.JOB_OPERATION_ARCHIVE,
+                syncthing=SlowRescanSyncthing(),
+            )
+
     def _run_transfer(
         self,
         root: Path,
@@ -85,6 +118,7 @@ class ClientJobHappyPathTests(unittest.TestCase):
         *,
         source_stage_free_bytes: int | None = None,
         content: bytes = b"archive-control-happy-path",
+        syncthing: CompleteSyncthing | None = None,
     ):
         source_root = root / "source"
         target_root = root / "target"
@@ -152,7 +186,7 @@ class ClientJobHappyPathTests(unittest.TestCase):
         target_qb.get_resource.side_effect = [
             None, None, resource, resource,
         ]
-        syncthing = CompleteSyncthing()
+        syncthing = syncthing or CompleteSyncthing()
         source = ClientJobExecutor(
             client_id=source_id,
             qbittorrent=source_qb,

tests/test_routes.py

@@ -62,6 +62,32 @@ class RouteDiscoveryTests(unittest.TestCase):
         self.assertEqual(routes[0].local_relative_path, "DownloadsSync")
         self.assertEqual(routes[0].state, route_pb2.ROUTE_STATE_DISCOVERED)
 
+    def test_discovery_accepts_a_safe_local_folder_override(self):
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            override = root / "qb/Sync"
+            override.mkdir(parents=True)
+            roots = RootMapping(
+                PurePosixPath("/var/syncthing"),
+                root / "sync",
+                ((PurePosixPath("/var/syncthing/DownloadsSync"), override),),
+            )
+            routes = discover_routes(
+                {"folders": [{
+                    "id": "DownloadsSync",
+                    "path": "~/DownloadsSync",
+                    "type": "sendreceive",
+                    "devices": [
+                        {"deviceID": "LOCAL"},
+                        {"deviceID": "ARCHIVE"},
+                    ],
+                }]},
+                "LOCAL",
+                roots,
+                True,
+            )
+        self.assertEqual([route.route_id for route in routes], ["DownloadsSync"])
+
 
 if __name__ == "__main__":
     unittest.main()

tests/test_transfer.py

@@ -11,6 +11,7 @@ from archive_clients.transfer import (
     FileMaterializer,
     TransferIntegrityError,
     canonical_message_json,
+    cleanup_partial_transfer,
     load_published_transfer,
     materialize_transfer,
     stage_transfer,
@@ -206,6 +207,33 @@ class TransferHappyPathTests(unittest.TestCase):
             os.stat(copied).st_blocks * 512, os.stat(copied).st_size
         )
 
+    def test_partial_cleanup_removes_only_journalled_temporary(self):
+        job_id = str(uuid4())
+        job_root = self.sync / ".archive-control/jobs" / job_id
+        destination = job_root / "payload/album/one.bin"
+        destination.parent.mkdir(parents=True)
+        operation_id = "interrupted-copy"
+        self.store.begin_file_operation(
+            operation_id,
+            job_id,
+            '{"destination":"payload/album/one.bin"}',
+        )
+        temporary = destination.with_name(
+            ".one.bin.archive-control-"
+            + hashlib.sha256(operation_id.encode()).hexdigest()[:16]
+            + ".tmp"
+        )
+        temporary.write_bytes(b"partial")
+        unrelated = destination.parent / "keep-me"
+        unrelated.write_bytes(b"unrelated")
+
+        cleanup_partial_transfer(
+            job_root, job_id=job_id, store=self.store
+        )
+
+        self.assertFalse(temporary.exists())
+        self.assertTrue(unrelated.is_file())
+
     def test_canonical_manifest_json_is_stable(self):
         manifest = self._manifest()
         first = canonical_message_json(manifest)