Gate a restore on every schema chain, not just the core one
Some checks failed
CI / naming (push) Successful in 4s
CI / frontend (push) Successful in 13s
CI / migrations-mysql (push) Failing after 15s
CI / backend (push) Failing after 4m56s

The compatibility gate read alembic_version and nothing else, but plugin schemas
move independently of core. Roll the code back to an older plugin revision and
the core head can be identical, so a dump carrying plugin tables the code has
never heard of passed the gate, loaded - destroying what was there - and only
then failed in 'flask plugin upgrade-all' with "Can't locate revision". A failed
restore that has already replaced what it was restoring over is the worst
outcome this feature can produce, and it was reachable from an ordinary rollback.

Backups now stamp the applied revision of every alembic_version_<plugin> table
alongside the core one, read from the dumped database the same way. restore.sh
checks each against that plugin's own versions directory and refuses before
touching anything, naming the plugin and the revision to deploy past. A plugin
in the dump that this build does not have at all is refused for the same reason.
The list endpoint's compatibility badge asks the same question, so the UI cannot
offer a set the restore will reject.

Sets taken before the stamp existed carry no dataplugins and keep their old
core-only verdict. Their plugin revisions are unknowable, and making every
existing set unrestorable to close this would be a worse trade than the gap.

Proven against a real stack, reproducing the exact case: core head identical,
endpoints data stamped at a revision the build does not have. The restore
refused, the live row was still there afterwards, and it stopped early enough
that the pre-restore safety copy was never needed. The same set with its true
revision restored normally.

Suite 2605 passed / 9 skipped, with the gate covered both ways plus the legacy
fallback.

Scope: product
This commit is contained in:
cproudlock
2026-09-06 18:02:39 -04:00
parent fa6b99bf91
commit e2705d732a
6 changed files with 246 additions and 6 deletions

View File

@@ -101,6 +101,47 @@ if ! grep -rqE "^revision = ['\"]${SET_DATAREV}['\"]" /app/migrations/versions 2
fi
log "data revision ${SET_DATAREV} is known to this code - restore may proceed"
# The same gate for every PLUGIN chain, because the core chain alone cannot
# answer this. Plugin schemas move independently of core: roll the code back to
# an older plugin revision and the core head can be identical while the dump
# carries plugin tables the code has never heard of. Gating on core only let
# that pass, load the dump, and then fail in 'flask plugin upgrade-all' with
# "Can't locate revision" and the data already replaced - a failed restore that
# has destroyed what it was restoring over.
#
# A manifest with no dataplugins is a legacy set taken before this was stamped.
# Its plugin revisions are unknowable, so it is allowed through on the core gate
# alone (the behaviour before this check existed) rather than being made
# unrestorable retroactively.
SET_DATAPLUGINS="$(python3 -c \
'import json,sys
try:
data = json.load(open(sys.argv[1])).get("dataplugins") or {}
except Exception:
data = {}
for name, rev in sorted(data.items()):
print(name + " " + rev)' \
"${SETDIR}/manifest.json" 2>/dev/null || echo "")"
if [ -z "${SET_DATAPLUGINS}" ]; then
log "manifest stamps no plugin revisions (legacy set) - core gate only"
else
while read -r plugin rev; do
[ -n "${plugin}" ] || continue
chain="/app/plugins/${plugin}/migrations/versions"
if [ ! -d "${chain}" ]; then
die "the backup carries data for plugin '${plugin}', which this build does not have - deploy a build that includes it before restoring"
fi
if ! grep -rqE "^revision = ['\"]${rev}['\"]" "${chain}" 2>/dev/null; then
log "plugin ${plugin} data revision: ${rev}"
die "plugin '${plugin}' data is at a revision this code does not have - deploy code at or past '${rev}' before restoring"
fi
log "plugin ${plugin} at ${rev} - known to this code"
done <<PLUGINREVS
${SET_DATAPLUGINS}
PLUGINREVS
fi
# Safety set FIRST. Everything below this point is destructive and none of it
# is transactional: the dump carries DROP TABLE per table, and the instance tree
# is replaced wholesale. Without a copy taken here, a restore of the wrong set -