Files
bvandeusenandClaude Opus 5.5 aab8581f20
CI / lint (push) Failing after 3s
CI / unit (push) Successful in 1m41s
CI / integration (push) Failing after 1m45s
CI / publish (push) Skipped
feat(backup): restore a same-version backup in place, history kept, all or nothing (#5815)
#5815 "Restore engine: in place for a same-version backup".

restore_in_place (steward/backup/restore.py) runs in one transaction:
1. It keeps aside what Steward records as it runs: ACTIVITY columns of rows
   that come back, and STATE_SETTINGS.
2. It drops every foreign key touching a replaced table.
3. It truncates the replaced tables and COPYs the backup's rows in
   unparsed, through asyncpg.
4. It puts the kept state back and moves sequences past their rows.
5. It puts right the rows whose parent didn't come back, by each key's own
   ON DELETE. A RESTRICT key that may be empty is cleared, so a run whose
   user is gone keeps its record.
6. It re-adds the keys, which checks every row.

A failure anywhere rolls the whole restore back. A backup made at another
schema is refused with NotSameSchema ("older": a full replace is the way;
"newer": update Steward).

The COPY-text reading moves into steward/backup/pgdata.py, shared by the
check and the restore. The integration tests stand in for pg_restore with
this database's own COPY output:
- settings come back while the audit log and the history of what remains
  stay;
- history of what didn't come back goes;
- a failure part-way changes nothing;
- a newer schema is refused.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-15 10:53:38 -04:00

186 lines
9.2 KiB
Python

"""Checking a backup against what is in place (steward/backup/check.py, #5823):
COPY text reads back exactly, rows are compared on their keys, secrets are
compared by what they open to and never shown, and what Steward records as
it runs is left out of the comparison."""
from datetime import datetime, timezone
from quart import render_template
from steward.backup import check
from steward.backup.check import (AreaCheck, CheckReport, RowDiff, TableCheck, _Opener, _Side,
compare_table, normalise, schema_of)
from steward.backup.pgdata import (copy_rows, script_blocks, script_from, script_tables,
unescape)
from steward.core.crypto import _make_fernet
from tests.test_style_tokens import hex_colours
KEY_A, KEY_B = "backup-key", "here-key"
def _seal(key: str, plain: str) -> str:
return "enc:v1:" + _make_fernet(key).encrypt(plain.encode()).decode()
def _compare(table, backup_cols, backup_rows, here_cols, here_rows, *, types=None, key=("id",),
keys=(KEY_A, KEY_B)):
return compare_table(table, _Side(backup_cols, backup_rows), _Side(here_cols, here_rows),
types=types or {}, key=list(key),
openers=(_Opener(keys[0]), _Opener(keys[1])), labels={})
def test_copy_text_reads_back_exactly():
assert unescape(b"\\N") is None
assert unescape(b"plain") == "plain"
assert unescape(b"a\\tb\\nc\\\\d") == "a\tb\nc\\d"
assert unescape(b"\\101\\x42") == "AB"
assert unescape("caf\u00e9".encode()) == "caf\u00e9"
assert copy_rows(b"1\tx\\ty\n2\t\\N\n") == [["1", "x\ty"], ["2", None]]
def test_a_restore_script_gives_each_tables_rows():
script = (b"SET client_encoding = 'UTF8';\n"
b'COPY public.app_settings (key, value_json, updated_at) FROM stdin;\n'
b'smtp.port\t587\t2026-10-15 14:00:00+00\n'
b'\\.\n\n'
b'COPY public.audit_events (id, "timestamp", action) FROM stdin;\n'
b'a1\t2026-10-15 14:00:00+00\tbackup.made\n'
b'\\.\n')
tables = script_tables(script)
assert tables["app_settings"] == (["key", "value_json", "updated_at"],
[["smtp.port", "587", "2026-10-15 14:00:00+00"]])
assert tables["audit_events"][0] == ["id", "timestamp", "action"]
# Handed on unparsed, a block is the COPY text as the dump holds it.
assert script_blocks(script)["app_settings"][1] == b"smtp.port\t587\t2026-10-15 14:00:00+00\n"
assert script_tables(script_from(script_blocks(script))) == tables
def test_an_instant_is_the_same_whatever_zone_prints_it():
a = normalise("2026-10-15 16:00:00+02", "timestamp with time zone")
b = normalise("2026-10-15 14:00:00+00", "timestamp with time zone")
assert a == b == datetime(2026, 10, 15, 14, tzinfo=timezone.utc)
assert normalise('{"b": 1, "a": 2}', "jsonb") == normalise('{"a":2,"b":1}', "jsonb")
def test_rows_are_compared_on_their_key():
cols = ["id", "name", "target"]
result = _compare("monitors", cols,
[["1", "web", "https://a"], ["2", "dns", "1.1.1.1"], ["3", "gone", "x"]],
cols,
[["1", "web", "https://a"], ["2", "dns", "9.9.9.9"], ["4", "new", "y"]])
assert result.same == 1
assert [(d.label, d.columns) for d in result.differ] == [("dns", ["target"])]
assert result.only_backup == ["gone"] and result.only_here == ["new"]
assert result.changes == 3
def test_a_setting_that_differs_shows_both_values():
cols = ["key", "value_json", "updated_at"]
result = _compare("app_settings", cols, [["smtp.port", "587", "2026-10-01 00:00:00+00"]],
cols, [["smtp.port", "25", "2026-10-15 00:00:00+00"]], key=("key",))
[d] = result.differ
assert (d.label, d.secret, d.backup_value, d.here_value) == ("smtp.port", False, "587", "25")
def test_a_secret_sealed_twice_matches_and_one_that_changed_is_never_shown():
cols = ["key", "value_json"]
same = _compare("app_settings", cols, [["smtp.password", f'"{_seal(KEY_A, "hunter2")}"']],
cols, [["smtp.password", f'"{_seal(KEY_B, "hunter2")}"']], key=("key",))
assert same.same == 1 and not same.differ
changed = _compare("app_settings", cols, [["smtp.password", f'"{_seal(KEY_A, "old")}"']],
cols, [["smtp.password", f'"{_seal(KEY_B, "new")}"']], key=("key",))
[d] = changed.differ
assert d.secret and d.backup_value is None and d.here_value is None
def test_a_backup_whose_key_cant_open_its_secrets_is_counted():
cols = ["key", "value_json"]
openers = (_Opener(KEY_A), _Opener(KEY_B))
compare_table("app_settings", _Side(cols, [["ldap.bind_password", f'"{_seal("other", "x")}"']]),
_Side(cols, []), types={}, key=["key"], openers=openers, labels={})
assert (openers[0].seen, openers[0].failed) == (1, 1)
def test_what_steward_records_as_it_runs_is_not_a_difference():
cols = ["id", "host_id", "token_hash", "last_seen_at", "agent_version", "updated_at"]
result = _compare("host_agent_registrations", cols,
[["r1", "h1", "abc", "2026-10-15 01:00:00+00", "1.19.0", "x"]], cols,
[["r1", "h1", "abc", "2026-10-15 14:00:00+00", "1.20.0", "y"]])
assert result.same == 1 and not result.differ
settings = ["key", "value_json"]
state = _compare("app_settings", settings, [["backup.history", "[]"]],
settings, [["backup.history", '[{"ok": true}]']], key=("key",))
assert state.same == 0 and state.changes == 0
def test_a_network_client_counts_only_once_someone_named_it():
cols = ["mac", "name", "note", "ip"]
result = _compare("unifi_known_clients", cols, [["aa", "printer", None, "10.0.0.5"]], cols,
[["aa", "printer", None, "10.0.0.9"], ["bb", None, None, "10.0.0.7"]],
key=("mac",))
assert result.same == 1 and result.changes == 0
def test_columns_one_side_lacks_are_noted_not_compared():
result = _compare("monitors", ["id", "name"], [["1", "web"]],
["id", "name", "retries"], [["1", "web", "3"]])
assert result.same == 1 and "retries" in result.note
def test_the_schema_is_same_older_or_newer():
known = {"0001", "0002", "0003"}
assert schema_of(["0003"], ["0003"], known) == "same"
assert schema_of(["0002"], ["0003"], known) == "older"
assert schema_of(["0009"], ["0003"], known) == "newer"
def test_this_steward_knows_its_own_revisions():
revisions = check.known_revisions()
assert len(revisions) > 10 and any(r.startswith("host_agent_") for r in revisions)
def _report(**over):
tables = [TableCheck("app_settings", same=12,
differ=[RowDiff("smtp.port", ["value_json"], False, "587", "25"),
RowDiff("smtp.password", ["value_json"], True)]),
]
hosts = [TableCheck("hosts", same=3, only_here=["nas"])]
fields = dict(name="steward-20261015T030000Z.tar",
manifest={"created_at": "2026-10-15T03:00:00+00:00", "steward_version": "2026.10.15"},
schema="same", backup_heads=["a"], here_heads=["a"], key_same=True, sealed=4,
unopened=0, areas=[AreaCheck("Settings", tables), AreaCheck("Hosts and agents", hosts)])
return CheckReport(**(fields | over))
def test_the_summary_says_whether_it_matches():
assert _report().summary() == "3 differences from what is in place: 2 in Settings, 1 in Hosts and agents."
same = _report(areas=[AreaCheck("Settings", [TableCheck("app_settings", same=12)])])
assert same.summary() == "Matches what is in place: 12 rows across 1 table." and same.ok
bad = _report(unopened=4)
assert not bad.ok and "doesn't open 4 of its 4 sealed values" in bad.summary()
assert _report().as_entry() == {"ok": True, "changes": 3, "rows": 18,
"summary": _report().summary()}
async def test_the_check_page_shows_what_differs_and_hides_secrets(app):
async with app.test_request_context("/settings/backups/check/", method="POST"):
page = await render_template("settings/backup_check.html", report=_report(), error="",
name="steward-20261015T030000Z.tar")
page = " ".join(page.split())
assert 'href="/settings/backups/"' in page # the trail back
assert "3 differences from what is in place" in page
assert "<strong>smtp.port</strong> differs: the backup has <code>587</code>, here it is <code>25</code>." in page
assert "<strong>smtp.password</strong> differs (a sealed value; not shown)." in page
assert "<strong>nas</strong> is only here: a restore removes it." in page
assert "The same as this Steward&#39;s." in page or "The same as this Steward's." in page
assert not hex_colours(page), hex_colours(page)
async def test_a_check_that_couldnt_run_says_why(app):
async with app.test_request_context("/settings/backups/check/", method="POST"):
page = await render_template("settings/backup_check.html", report=None,
error="Reading x refused: HTTP 403 AccessDenied.", name="x")
assert "x couldn&#39;t be checked." in page or "x couldn't be checked." in page
assert "AccessDenied" in page