Files
cproudlock 8fee09e748 Validate what a form accepts, and keep a diff to one asset
The validation tail from the 2026-09-03 run. Each is small; together they are
the difference between a field that checks its input and one that looks like it
does.

Subnets accepted any string containing a slash, so 'notanip/xx' was stored -
and _devices_in_subnet then swallowed the ValueError from ip_network and
returned an empty device list, so the subnet looked real and permanently empty.
It is parsed with the same library that has to read it later, a prefix length is
required (ipaddress reads a bare 10.1.1.0 as a /32, which is legal and never
what somebody typing a subnet meant), and the stored form is normalised so
10.1.1.5/24 and 10.1.1.0/24 cannot become two rows nobody can tell apart.

VLAN numbers had no range check: 99999 and -5 were stored. Worse, a non-numeric
value reached a duplicate probe that compares against an integer column, which
MySQL coerces to 0 - so 'xyz' matched an existing VLAN 0 and the 409 named a row
that had nothing to do with it. 1-4094 per 802.1Q, checked before the probe.
Deleting a VLAN that subnets still reference now answers 409 like every sibling
in-use guard, rather than 400, which told the caller their request was malformed
when the request was fine and the state was the problem.

An appversion sha256 was stored raw, so an over-long value was truncated to the
column width and came back as 64 plausible hex characters - indistinguishable
from a real digest, and wrong. A digest is a claim the delivery path checks
against, so it is 64 hex characters or it is refused. The releasedate goes
through the strict parser for the same reason it did everywhere else.

PUT /api/settings/<key> ignored the declared valuetype. "A6-not-a-number" went
into an integer key verbatim and every consumer then read 0, because
get_typed_value swallows the ValueError - the setting looked set and the
behaviour was the default.

/api/maplevels told a caller who omitted the level name that 'buildingname and
buildingid are required'; buildingname is a field that handler never reads. It
now names the fields it actually wants.

Restoring a guide revision that is already current writes nothing, because
save() dedups on the content hash - and the response still said "Restored
version N as a new draft version", describing a revision that does not exist
while the unchanged revision list disagreed.

GET /api/backups/revisions/<id>/diff?against=<id> accepted any revision id, so
pointing it at another asset's revision rendered that asset's configuration as
the "before" side of this asset's history: a diff that reads as a change on this
machine and describes a different one. Same asset and same kind, or refused.
The default branch was always asset-scoped; only the explicit one was not.

Two frontend fixes: the per-page selector rendered blank on any list defaulting
to a size not in [10,20,50,100] (network devices uses 25), so the operator could
not see the current page size; and an emptied PC alias was dropped from the
payload instead of sent, so it could not be cleared - the form came back showing
the alias that had just been deleted.

Twenty-two tests. The backups guard verified by planting.

Scope: product
2026-09-03 16:13:59 -04:00

410 lines
16 KiB
Python

"""Backups API.
Read + download only. Revisions are created exclusively through the ADR-006
collector endpoint (POST /api/collector/backups), so there is deliberately no
create route here: a hand-posted "backup" that never came off a machine would
poison the history this feature exists to provide.
Downloads differ by storage backend. 'shopdb' kinds re-render from the stored
projection, which is what lets one revision produce either .reg dialect. 'share'
kinds are not streamed - ShopDB returns the UNC path for the tech to open,
because requiring the app server to mount the SFLD share would turn a
permissions slip into an unexplained empty download.
"""
from datetime import timezone
from zoneinfo import ZoneInfo
from flask import Blueprint, current_app, request, Response
from flask_jwt_extended import jwt_required
from shopdb.api import (
db, Asset,
success_response, error_response, ErrorCodes,
require_permission,
)
from ..models import BackupRevision
from ..models.backup import _utciso
from ..services.registry import REGISTRY, getkind
from ..services.retention import samesource
backups_bp = Blueprint('backups', __name__)
_DEFAULTTZ = 'America/New_York'
def _sitezone():
"""Site-configured IANA zone (settings key site_timezone).
Same lookup the notifications plugin uses. A stored timestamp is UTC, so
anything rendered server-side has to be converted or it shows the wrong
wall clock for the site - four hours out at the reference site.
"""
from shopdb.api import Setting
row = Setting.query.filter_by(key='site_timezone').first()
name = row.value if row and row.value else _DEFAULTTZ
try:
return ZoneInfo(name)
except Exception:
return ZoneInfo(_DEFAULTTZ)
def _label(revision):
"""Panel list title: kind-agnostic, readable at a glance.
Rendered in the SITE zone, not UTC: this string is baked server-side and
the client cannot correct it afterwards.
"""
when = revision.collectedat or revision.createdat
if not when:
return 'unknown time'
local = when.replace(tzinfo=timezone.utc).astimezone(_sitezone())
stamp = local.strftime('%Y-%m-%d %H:%M')
if revision.sourcefilename:
return '{} - {}'.format(stamp, revision.sourcefilename)
return stamp
def _summary(revision, islatest=False):
data = revision.to_dict()
data['label'] = _label(revision)
data['islatest'] = islatest
kind = getkind(revision.backupkind)
data['formats'] = kind.formats() if kind else []
return data
@backups_bp.route('/kinds', methods=['GET'])
@require_permission('backups.view')
def list_kinds():
"""Registered backup kinds, for UI that needs to enumerate them."""
return success_response([
{
'key': kind.key,
'displayname': kind.displayname,
'storagebackend': kind.storagebackend,
'assettypes': list(kind.assettypes),
'formats': kind.formats(),
}
for kind in REGISTRY.values()
])
@backups_bp.route('/asset/<int:assetid>', methods=['GET'])
@require_permission('backups.view')
def asset_revisions(assetid):
"""Revision history for one asset, newest first.
This is the asset-panel endpoint. Optional ?kind= narrows to a single kind,
which is how each per-kind panel scopes itself.
"""
asset = db.session.get(Asset, assetid)
if asset is None:
return error_response(ErrorCodes.NOT_FOUND, 'Asset not found', http_code=404)
query = db.session.query(BackupRevision).filter(
BackupRevision.assetid == assetid)
kindkey = (request.args.get('kind') or '').strip().lower()
if kindkey:
if getkind(kindkey) is None:
return error_response(ErrorCodes.VALIDATION_ERROR,
'Unknown kind: {}'.format(kindkey))
query = query.filter(BackupRevision.backupkind == kindkey)
revisions = query.order_by(BackupRevision.backuprevisionid.desc()).all()
# "Latest" is per kind AND per source chain. Per kind alone was wrong on
# the case this exists for: NTLARS settings describe a machine but are read
# off the controlling PC, so a machine with two reporting PCs has two
# independent chains of the same kind. Marking only the highest revision id
# meant the second PC's current revision was badged as history and lost its
# "last verified" stamp - the machine looked like it had one source that
# kept re-reporting, rather than two.
#
# sourcehostname is the chain: it records which PC produced the revision.
# A revision with none (an older row, or a share-drop with no host) forms
# its own chain under the empty key, which keeps the per-kind behaviour for
# everything that has only ever had one source.
seen = set()
out = []
for revision in revisions:
chain = (revision.backupkind,
(revision.sourcehostname or '').strip().lower())
islatest = chain not in seen
seen.add(chain)
out.append(_summary(revision, islatest=islatest))
return success_response(out)
@backups_bp.route('/asset/<int:assetid>/info', methods=['GET'])
@require_permission('backups.view')
def asset_info(assetid):
"""A kind's 'at a glance' card, built from its LATEST revision.
Generic across kinds: the requested kind owns both the panel declaration
and the payload (see BackupKind.infopanel / buildinfo), so a successor to
NTLARS/DNC gets its own card without a new endpoint.
Empty when the machine has no revision of that kind yet, which the generic
renderer turns into the panel's empty text.
"""
kindkey = (request.args.get('kind') or 'ntlars').strip().lower()
kind = getkind(kindkey)
if kind is None:
return error_response(ErrorCodes.VALIDATION_ERROR,
'Unknown kind: {}'.format(kindkey))
revision = (db.session.query(BackupRevision)
.filter(BackupRevision.assetid == assetid,
BackupRevision.backupkind == kindkey)
.order_by(BackupRevision.backuprevisionid.desc())
.first())
if revision is None:
return success_response({'fields': [], 'sectioncount': 0})
try:
projection = revision.payload
except ValueError:
# payload deliberately raises on unreadable payloadjson so a corrupt row
# is named rather than mistaken for an opaque one. That is right for a
# download; for an at-a-glance card it would take the whole asset page
# down over one bad row, so the card degrades to empty instead.
current_app.logger.warning(
'backuprevision %s has unreadable payloadjson; info card empty',
revision.backuprevisionid)
return success_response({'fields': [], 'sectioncount': 0})
data = kind.buildinfo(
projection, assetid,
partmarkertypes=current_app.config.get('BACKUPS_PARTMARKER_TYPES'),
# A kind that diffs against its own previous revision needs to know
# which revision it is looking at. Passed generically rather than
# queried inside the kind, so the endpoint stays the single place that
# decides "latest".
revisionid=revision.backuprevisionid,
# The card names the PC a snapshot came from, and chains its diff on the
# same value. Passed from the column rather than read out of the stored
# manifest, which carries its own copy that can disagree.
sourcehostname=revision.sourcehostname,
# Where the snapshot actually is. Read off the revision rather than
# composed from the configured share root, so a snapshot written before
# the layout changed still points at where it was really put.
sharepath=revision.sharepath)
data['backuprevisionid'] = revision.backuprevisionid
data['collectedat'] = _utciso(revision.collectedat)
return success_response(data)
@backups_bp.route('/revisions/<int:backuprevisionid>', methods=['GET'])
@require_permission('backups.view')
def get_revision(backuprevisionid):
"""One revision, including its parsed payload where there is one."""
revision = db.session.get(BackupRevision, backuprevisionid)
if revision is None:
return error_response(ErrorCodes.NOT_FOUND, 'Revision not found',
http_code=404)
data = revision.to_dict(includepayload=True)
data['label'] = _label(revision)
kind = getkind(revision.backupkind)
data['formats'] = kind.formats() if kind else []
data['displayname'] = kind.displayname if kind else revision.backupkind
return success_response(data)
@backups_bp.route('/revisions/<int:backuprevisionid>/download', methods=['GET'])
@require_permission('backups.download')
def download_revision(backuprevisionid):
"""Render a revision back to its native file format.
?format=ntlars (default for NTLARS) omits WOW6432Node - the form the
NTLARS Load... button expects
?format=wow6432node includes WOW6432Node - for `reg import` on 64-bit
"""
revision = db.session.get(BackupRevision, backuprevisionid)
if revision is None:
return error_response(ErrorCodes.NOT_FOUND, 'Revision not found',
http_code=404)
kind = getkind(revision.backupkind)
if kind is None:
return error_response(ErrorCodes.VALIDATION_ERROR,
'Unknown kind: {}'.format(revision.backupkind))
if revision.storagebackend == 'share':
# Not an error: the file exists, ShopDB just is not the one serving it.
return success_response({
'storagebackend': 'share',
'sharepath': revision.sharepath,
'sourcefilename': revision.sourcefilename,
'message': 'This backup lives on the SFLD share. Open the path directly.',
})
formats = kind.formats()
formatid = (request.args.get('format') or '').strip().lower()
if not formatid:
formatid = formats[0]['id'] if formats else ''
if not any(f['id'] == formatid for f in formats):
return error_response(
ErrorCodes.VALIDATION_ERROR,
'Unknown format {!r} for kind {}'.format(formatid, kind.key))
asset = db.session.get(Asset, revision.assetid)
assetnumber = asset.assetnumber if asset else str(revision.assetid)
when = revision.collectedat or revision.createdat
comments = [
'{} backup from ShopDB'.format(kind.displayname),
'Asset: {}'.format(assetnumber),
'Captured: {}'.format(when.strftime('%Y-%m-%d %H:%M:%S') if when else 'unknown'),
'Source PC: {}'.format(revision.sourcehostname or 'unknown'),
'Revision: {} ({})'.format(revision.backuprevisionid,
(revision.contenthash or '')[:12]),
]
try:
raw, ext, mimetype = kind.render(revision.payload, formatid,
comments=comments)
except ValueError as exc:
return error_response(ErrorCodes.VALIDATION_ERROR, str(exc))
# Named for the MACHINE, not the revision or the kind: a tech restoring bay
# 3204 wants 3204.reg, matching how the per-machine backups on the share
# have always been named. sourcefilename is deliberately not reused here -
# a seeded revision carries "3204.reg" already and appending an extension
# to it produced "3204.reg.reg".
filename = '{}{}'.format(assetnumber, ext)
if formatid == 'wow6432node':
# The two dialects must not collide in a downloads folder, and the
# suffix says which one will import correctly outside NTLARS.
filename = '{}-wow6432node{}'.format(assetnumber, ext)
return Response(
raw,
mimetype=mimetype,
headers={'Content-Disposition': 'attachment; filename="{}"'.format(filename)},
)
@backups_bp.route('/revisions/<int:backuprevisionid>/diff', methods=['GET'])
@require_permission('backups.view')
def diff_revision(backuprevisionid):
"""Diff two revisions of the same asset+kind.
?against=<id> picks the comparison revision; default is the immediately
preceding one, which answers "what changed" without the user choosing.
"""
revision = db.session.get(BackupRevision, backuprevisionid)
if revision is None:
return error_response(ErrorCodes.NOT_FOUND, 'Revision not found',
http_code=404)
if revision.storagebackend != 'shopdb':
return error_response(
ErrorCodes.VALIDATION_ERROR,
'Kind {} stores opaque files and cannot be diffed'.format(
revision.backupkind))
againstid = request.args.get('against', type=int)
if againstid:
other = db.session.get(BackupRevision, againstid)
# It has to belong to the SAME asset. Any revision id was accepted, so
# ?against=<another asset's revision> rendered that asset's
# configuration values as the "before" side of this asset's history -
# a diff that reads as a change on this machine and describes a
# different one. The default branch below has always been asset-scoped;
# only the explicit one was not.
if other is not None and other.assetid != revision.assetid:
return error_response(
ErrorCodes.VALIDATION_ERROR,
'Revision {} belongs to a different asset and cannot be '
'compared against this one'.format(againstid))
if other is not None and other.backupkind != revision.backupkind:
return error_response(
ErrorCodes.VALIDATION_ERROR,
'Revision {} is a {} backup and this one is {} - there is '
'nothing to compare'.format(againstid, other.backupkind,
revision.backupkind))
else:
# The previous revision FROM THE SAME PC. Several PCs can share a
# machine number, and diffing across them reported one device's COM port
# as a change on another, which is not a change at all.
other = (db.session.query(BackupRevision)
.filter(BackupRevision.assetid == revision.assetid,
BackupRevision.backupkind == revision.backupkind,
samesource(revision.sourcehostname),
BackupRevision.backuprevisionid < revision.backuprevisionid)
.order_by(BackupRevision.backuprevisionid.desc())
.first())
if other is None:
return success_response({
'backuprevisionid': revision.backuprevisionid,
'againstid': None,
'changes': [],
'message': 'No earlier revision to compare against.',
})
changes = _diffprojections(other.payload, revision.payload)
return success_response({
'backuprevisionid': revision.backuprevisionid,
'againstid': other.backuprevisionid,
'changes': changes,
'changecount': len(changes),
})
def _flatten(projection):
"""{(subkey, valuename): (type, data)} from a stored projection."""
flat = {}
for key in (projection or {}).get('keys', []):
path = key.get('path', '')
for name, entry in (key.get('values') or {}).items():
flat[(path, name)] = (entry.get('type'), entry.get('data'))
return flat
def _diffprojections(old, new):
oldflat = _flatten(old)
newflat = _flatten(new)
changes = []
for ref in sorted(set(oldflat) | set(newflat)):
path, name = ref
before = oldflat.get(ref)
after = newflat.get(ref)
if before == after:
continue
if before is None:
change = 'added'
elif after is None:
change = 'removed'
else:
change = 'changed'
changes.append({
'keypath': path,
'valuename': name,
'change': change,
'before': None if before is None else before[1],
'after': None if after is None else after[1],
'beforetype': None if before is None else before[0],
'aftertype': None if after is None else after[0],
})
return changes
@backups_bp.route('/dashboard/stale', methods=['GET'])
@require_permission('backups.view')
def dashboard_stale():
"""Chains whose backup has stopped running.
Thin: the rule lives in services/staleness.py, where it is testable without
an auth layer in the way. Keyed on the last CONFIRMED check, never on the
last change - dedup means an unchanged config writes no revision, so a card
keyed on revision age would flag most of a healthy fleet.
"""
from ..services.staleness import stalechains
return success_response(stalechains())