Merge remote-tracking branch 'origin/main' into codex/sre-typed-automation-20260715

scripts/ops/deploy-signoz-metadata-toolchain.sh

@@ -395,7 +395,14 @@ mkdir -m 0700 "${RECEIPT_DIR}" "${CANDIDATE_DIR}"
 : > "${RECEIPT_LOG}"
 chmod 0600 "${RECEIPT_LOG}"
 
-exec 8>>/tmp/awoooi-signoz-backup-operation.lock
+OPERATION_LOCK=/tmp/awoooi-signoz-backup-operation.lock
+[ -f "${OPERATION_LOCK}" ] && [ ! -L "${OPERATION_LOCK}" ]
+# The shared lock is intentionally owned by the unprivileged backup account.
+# Open the existing inode read-only so Linux fs.protected_regular=2 cannot
+# reject a root O_CREAT/O_APPEND open in sticky /tmp. flock(2) still provides
+# the same exclusive inter-process lock without changing file contents,
+# ownership, or mode. Missing/unsafe lock state fails closed.
+exec 8<"${OPERATION_LOCK}"
 flock -n 8
 active_operations="$(
   (pgrep -af '(^|/|[[:space:]])(backup-signoz|clickhouse-native-backup|clickhouse-native-restore-drill|run-signoz-backup-canary)[.]sh([[:space:]]|$)' || true) \

scripts/ops/tests/test_signoz_metadata_toolchain_deploy.py

@@ -165,6 +165,15 @@ def test_deployer_has_no_active_pointer_or_destructive_fallback() -> None:
     assert "ln -s" not in source
 
 
+def test_shared_operation_lock_is_existing_read_only_and_fail_closed() -> None:
+    source = DEPLOYER.read_text(encoding="utf-8")
+    assert "OPERATION_LOCK=/tmp/awoooi-signoz-backup-operation.lock" in source
+    assert '[ -f "${OPERATION_LOCK}" ] && [ ! -L "${OPERATION_LOCK}" ]' in source
+    assert 'exec 8<"${OPERATION_LOCK}"' in source
+    assert "flock -n 8" in source
+    assert "exec 8>>/tmp/awoooi-signoz-backup-operation.lock" not in source
+
+
 def test_exact_five_file_supply_chain_and_modes_are_fixed() -> None:
     source = DEPLOYER.read_text(encoding="utf-8")
     for path in (

scripts/reboot-recovery/signoz-canonical-route-preflight.py

@@ -865,7 +865,7 @@ def evaluate(root: Path, contract_path: Path) -> dict[str, Any]:
         == len(METADATA_EXPORT_SOURCE_PATHS)
         and set(declared_metadata_export_hashes) == set(METADATA_EXPORT_SOURCE_PATHS)
         and declared_metadata_export_hashes == actual_metadata_export_hashes
-        and metadata_export_source_contract.get("test_terminal") == "19_passed"
+        and metadata_export_source_contract.get("test_terminal") == "20_passed"
     )
     offsite_backup = pending_verifiers.get("signoz_backup_offsite_dr", {})
     failed_artifacts = pending_verifiers.get(

scripts/reboot-recovery/tests/test_signoz_canonical_route_migration.py

@@ -877,7 +877,7 @@ def test_post_release_verifier_and_native_backup_close_with_remaining_gaps() ->
         field: hashlib.sha256((ROOT / relative).read_bytes()).hexdigest()
         for field, relative in metadata_paths.items()
     }
-    assert source_contract["test_terminal"] == "19_passed"
+    assert source_contract["test_terminal"] == "20_passed"
     assert source_contract["raw_sqlite_read"] is False
     assert source_contract["raw_volume_read"] is False
     assert source_contract["secret_value_persistence"] is False