Distinguish a missing snapshot from an unreadable one

os.path.isdir() returns False both when a snapshot directory does not exist and
when it cannot be stat'd. Staging aborted either way -- correct -- but reported
every case as "has no snapshot", which sends you hunting for a snapshot that is
sitting right there.

Found while dry-running the planner against a real recursive snapshot of Tap:
running as a non-root user, /mnt/Tap/apps/paperless/data is mode 0700 and the
probe reported "has no snapshot" when `zfs list` showed the snapshot present.
Middleware runs as root so this would not fire in production, but a backup
system must not misreport why it failed.

plan_staging now takes a probe() that classifies the path as ok / missing /
unreadable, and the error names which.

Dry-run results against Tap (250 datasets, real `zfs snapshot -r`):
- 170 mounted filesystems under /mnt/Tap/, and the plan produces exactly 170
  descendant mounts -- no omissions
- 18 legacy-mountpoint datasets reported as skipped, never dropped silently
- Tap/ix-apps mounts at /mnt/.ix-apps, correctly outside the backup path
- parent snapshot exposes 0 entries under /apps; the staged sources expose 71,
  and lidarr/config resolves with lidarr.db present
- snapshot_tree_names() identifies all 250; deleting only the parent (what stock
  does) leaves 249 orphans, and the sweep clears them
This commit is contained in:
flan
2026-07-12 22:04:56 +00:00
parent bb26edf351
commit a80de88078
2 changed files with 52 additions and 12 deletions
+30 -5
View File
@@ -56,6 +56,7 @@ from __future__ import annotations
import contextlib
import os
import stat
import subprocess
__all__ = [
@@ -155,8 +156,25 @@ def snapshot_tree_names(snapshot: str, all_names) -> list[str]:
]
def _probe_snapdir(path):
"""Classify a snapshot directory: ``ok``, ``missing``, or why it is unusable.
``os.path.isdir()`` collapses "does not exist" and "cannot stat" into the
same ``False``, so an EACCES would report itself as "has no snapshot" and
send someone hunting for a snapshot that is sitting right there. Both cases
still abort the backup -- but it has to say which one.
"""
try:
st = os.stat(path)
except FileNotFoundError:
return "missing"
except OSError as e:
return f"cannot be read ({e.strerror})"
return "ok" if stat.S_ISDIR(st.st_mode) else "is not a directory"
def plan_staging(base_dataset, base_mountpoint, path, snapshot_name, datasets,
staging_root, isdir=os.path.isdir):
staging_root, probe=_probe_snapdir):
"""Compute the bind-mount plan for staging a nested tree. Pure function.
``datasets`` is a list of dicts shaped like ``zfs.dataset.query`` results:
@@ -216,11 +234,18 @@ def plan_staging(base_dataset, base_mountpoint, path, snapshot_name, datasets,
continue
src = snapdir(mp)
if not isdir(src):
# The recursive snapshot should have covered every descendant. If it
# did not, this dataset's data would be silently omitted. Refuse.
status = probe(src)
if status != "ok":
# Either the recursive snapshot missed this dataset, or we cannot read
# it. Either way its data would be silently omitted. Refuse -- but say
# WHICH, because "no snapshot" and "permission denied" send you to
# completely different places.
detail = (
f"has no snapshot {snapshot_name!r}" if status == "missing"
else f"snapshot {snapshot_name!r} {status}"
)
raise StagingError(
f"dataset {name!r} has no snapshot {snapshot_name!r} at {src!r}; "
f"dataset {name!r} {detail} at {src!r}; "
f"refusing to back up an incomplete tree"
)