Reject unfinished or truncated Forgejo backup archives before encryption
All checks were successful
CI Smoke / host-smoke (push) Successful in 0s
CI Smoke / container-smoke (push) Successful in 2s

Assistant: codex
Assistant-Model: gpt-6-astra
Assistant-Session: 01a06ecb-456a-71c2-b41e-0755d336e883
This commit is contained in:
codex 2026-09-05 22:10:22 +02:00
parent 6124a15c24
commit d0c7249a0d
3 changed files with 98 additions and 93 deletions

View file

@ -0,0 +1,72 @@
#!/usr/bin/env python3
"""Capture only a completed, byte-identical and CRC-verified Forgejo ZIP."""
import argparse
import hashlib
from pathlib import Path
import secrets
import subprocess
import time
import zipfile
def validate_archive(path):
with zipfile.ZipFile(path) as archive:
names = set(archive.namelist())
if 'forgejo-db.sql' not in names or not any(n.startswith('repos/') for n in names):
raise ValueError('required archive content missing')
if archive.testzip() is not None:
raise ValueError('archive checksum failure')
def capture(namespace, pod, destination):
k = ['kubectl', 'exec', '--request-timeout=120s', '-n', namespace, pod, '-c', 'gitea', '--']
def call(args, timeout=150):
r = subprocess.run(k + args, capture_output=True, timeout=timeout)
if r.returncode: raise ValueError('capture command failed')
return r.stdout
remote = '/tmp/wp0029-backup-' + secrets.token_hex(12)
completed = False
try:
# Completion belongs to this exact process, not a global pgrep or file existence.
command = f'umask 077; forgejo dump -f {remote}.zip >{remote}.log 2>&1; result=$?; printf "%s" "$result" >{remote}.exit'
call(['sh','-c', 'nohup sh -c "$1" >/dev/null 2>&1 </dev/null &', 'sh', command])
for _ in range(180):
r = subprocess.run(k + ['cat',remote+'.exit'], capture_output=True, timeout=30)
if r.returncode == 0:
completed = True
if r.stdout.strip() != b'0': raise ValueError('dump process failed')
break
time.sleep(10)
else: raise ValueError('dump completion timeout')
size = int(call(['stat','-c%s',remote+'.zip']).strip())
expected = call(['sha256sum',remote+'.zip']).split()[0].decode()
if size <= 0: raise ValueError('empty archive')
chunk = 4*1024*1024
with destination.open('xb') as out:
destination.chmod(0o600)
for index in range((size+chunk-1)//chunk):
piece = call(['dd','if='+remote+'.zip','bs='+str(chunk),'skip='+str(index),'count=1'])
if len(piece) != min(chunk,size-index*chunk): raise ValueError('short archive chunk')
out.write(piece)
actual = hashlib.file_digest(destination.open('rb'),'sha256').hexdigest()
if actual != expected: raise ValueError('archive transfer mismatch')
validate_archive(destination)
finally:
# Do not remove a file while a timed-out producer may still be writing it.
if completed:
call(['rm','-f',remote+'.zip',remote+'.log',remote+'.exit'])
def main():
p=argparse.ArgumentParser(description=__doc__)
p.add_argument('--namespace',required=True); p.add_argument('--pod',required=True)
p.add_argument('--output',required=True,type=Path)
a=p.parse_args()
try: capture(a.namespace,a.pod,a.output)
except Exception:
print('ERROR: Forgejo archive capture or integrity validation failed')
return 1
print('Forgejo archive capture and integrity verified')
return 0
if __name__=='__main__': raise SystemExit(main())

View file

@ -0,0 +1,24 @@
import sys
import tempfile
import unittest
from pathlib import Path
from zipfile import ZipFile, BadZipFile
sys.path.insert(0,str(Path(__file__).resolve().parents[1]/'scripts'))
from capture_forgejo_archive import validate_archive
class ArchiveIntegrity(unittest.TestCase):
def test_complete_archive(self):
with tempfile.TemporaryDirectory() as d:
p=Path(d)/'backup.zip'
with ZipFile(p,'w') as z:
z.writestr('forgejo-db.sql','SELECT 1;')
z.writestr('repos/example.git/HEAD','ref: refs/heads/main')
validate_archive(p)
p.write_bytes(p.read_bytes()[:-22])
with self.assertRaises(BadZipFile): validate_archive(p)
def test_missing_database_rejected(self):
with tempfile.TemporaryDirectory() as d:
p=Path(d)/'backup.zip'
with ZipFile(p,'w') as z: z.writestr('repos/example.git/HEAD','x')
with self.assertRaises(ValueError): validate_archive(p)

View file

@ -87,99 +87,8 @@ ok "forgejo-db" "${FORGEJO_DB_NAMESPACE}/${FORGEJO_DB_POD}"
# 1. Application dump (repos, packages, attachments, LFS, avatars)
ok "forgejo dump" "running in pod…"
DUMP_REMOTE="/tmp/forgejo-backup-${TS}.zip"
# Long dumps drop the kubectl exec websocket if attached to the process; run in-pod
# background and poll for the archive file instead.
kubectl exec --request-timeout=0 -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
sh -eu -c "rm -f '${DUMP_REMOTE}'; nohup forgejo dump -f '${DUMP_REMOTE}' >/tmp/forgejo-dump.log 2>&1 &"
dump_wait=0
while [[ "${dump_wait}" -lt 1800 ]]; do
if kubectl exec --request-timeout=30 -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
test -s "${DUMP_REMOTE}" 2>/dev/null; then
if kubectl exec --request-timeout=30 -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
sh -c "pgrep -f 'forgejo dump' >/dev/null"; then
sleep 10
dump_wait=$((dump_wait + 10))
continue
fi
break
fi
sleep 10
dump_wait=$((dump_wait + 10))
done
if ! kubectl exec --request-timeout=30 -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
test -s "${DUMP_REMOTE}" 2>/dev/null; then
bad "forgejo dump" "timed out or empty archive (see /tmp/forgejo-dump.log in pod)"
exit 1
fi
# Large dumps (~700M) fail via stdout stream or single kubectl cp — use chunks.
dump_size="$(kubectl exec -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
stat -c%s "${DUMP_REMOTE}" 2>/dev/null || echo 0)"
refresh_prod_pod() {
PROD_POD="$(kubectl get pods -n "${FORGEJO_NAMESPACE}" \
-l "app.kubernetes.io/instance=${FORGEJO_RELEASE}" \
-o jsonpath='{.items[0].metadata.name}' 2>/dev/null || true)"
}
copy_from_pod() {
local remote_path="$1" local_path="$2" attempt
for attempt in 1 2 3 4 5; do
refresh_prod_pod
if [[ -z "${PROD_POD}" ]]; then
sleep $((attempt * 5))
continue
fi
if kubectl cp -n "${FORGEJO_NAMESPACE}" -c gitea \
"${FORGEJO_NAMESPACE}/${PROD_POD}:${remote_path}" "${local_path}" 2>/dev/null; then
return 0
fi
if kubectl exec --request-timeout=120 -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
cat "${remote_path}" > "${local_path}" 2>/dev/null && [[ -s "${local_path}" ]]; then
return 0
fi
sleep $((attempt * 5))
done
return 1
}
if [[ "${dump_size}" -lt 52428800 ]]; then
copy_from_pod "${DUMP_REMOTE}" "${DUMP_PLAIN}" || exit 1
else
ok "forgejo dump" "chunked copy (${dump_size} bytes)…"
chunk_dir="$(mktemp -d)"
refresh_prod_pod
parts="$(kubectl exec --request-timeout=120 -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
sh -c "rm -f /tmp/forgejo-backup-part-*; split -b 4m '${DUMP_REMOTE}' /tmp/forgejo-backup-part- && ls /tmp/forgejo-backup-part-*")"
for part in ${parts}; do
base="$(basename "${part}")"
local_chunk="${chunk_dir}/${base}"
if [[ -f "${local_chunk}" && -s "${local_chunk}" ]]; then
ok "forgejo dump" "reuse cached chunk ${base}"
continue
fi
chunk_ok=0
for chunk_try in 1 2 3 4 5 6 7 8; do
if copy_from_pod "${part}" "${local_chunk}"; then
chunk_ok=1
ok "forgejo dump" "chunk ${base} (${chunk_try})"
break
fi
sleep $((chunk_try * 3))
done
if [[ "${chunk_ok}" -ne 1 ]]; then
bad "forgejo dump" "chunk copy failed: ${base}"
exit 1
fi
done
cat "${chunk_dir}"/forgejo-backup-part-* > "${DUMP_PLAIN}"
rm -rf "${chunk_dir}"
kubectl exec -n "${FORGEJO_NAMESPACE}" "${PROD_POD}" -c gitea -- \
rm -f /tmp/forgejo-backup-part-* "${DUMP_REMOTE}" || true
fi
if [[ ! -f "${DUMP_PLAIN}" || ! -s "${DUMP_PLAIN}" ]]; then
bad "forgejo dump" "failed to copy dump from pod"
exit 1
fi
python3 "${ROOT}/scripts/capture_forgejo_archive.py" \
--namespace "${FORGEJO_NAMESPACE}" --pod "${PROD_POD}" --output "${DUMP_PLAIN}"
ok "forgejo dump" "$(du -h "${DUMP_PLAIN}" | awk '{print $1}') $(basename "${DUMP_PLAIN}")"
# 2. PostgreSQL logical dump (plain SQL to stdout; CNPG root FS is read-only)