fix(agent): preserve log consumer readback on trace failure
Some checks failed
CD Pipeline / workflow-shape (push) Successful in 0s
CD Pipeline / cancel-stale-cd (push) Has been skipped
CD Pipeline / tests (push) Successful in 44s
CD Pipeline / post-deploy-checks (push) Has been cancelled
CD Pipeline / build-and-deploy (push) Has been cancelled

This commit is contained in:
Your Name
2026-07-01 18:27:59 +08:00
parent d873a3258c
commit 7aa9cb66ec
2 changed files with 94 additions and 6 deletions

View File

@@ -3249,6 +3249,9 @@ def _attach_runtime_receipt_readback(
log_consumer_blockers = log_consumer.get("active_blockers")
if not isinstance(log_consumer_blockers, list):
log_consumer_blockers = []
log_consumer_dispatch_ledger_count = _int_value(
log_consumer_rollups.get("dispatch_ledger_row_count")
)
operation_counts = (readback.get("ansible_operations") or {}).get("counts")
if not isinstance(operation_counts, Mapping):
operation_counts = {}
@@ -3373,11 +3376,17 @@ def _attach_runtime_receipt_readback(
)
),
"live_log_controlled_writeback_dispatch_count": _int_value(
log_dispatch_summary.get("total")
max(
_int_value(log_dispatch_summary.get("total")),
log_consumer_dispatch_ledger_count,
)
),
"live_log_controlled_writeback_recent_dispatch_count": _int_value(
log_dispatch_summary.get("recent")
),
"live_log_controlled_writeback_consumer_dispatch_ledger_count": (
log_consumer_dispatch_ledger_count
),
"live_log_controlled_writeback_consumer_binding_count": _int_value(
log_consumer_rollups.get("consumer_binding_count")
),
@@ -3834,6 +3843,7 @@ async def load_ai_agent_autonomous_runtime_receipt_readback(
"lookback_hours": max(1, int(lookback_hours or _DEFAULT_LOOKBACK_HOURS)),
"limit": max(1, int(limit or 20)),
}
log_controlled_writeback_consumer: dict[str, Any] | None = None
try:
async with get_db_context(project_id) as db:
await db.execute(text("SET LOCAL statement_timeout = '5000ms'"))
@@ -3932,24 +3942,30 @@ async def load_ai_agent_autonomous_runtime_receipt_readback(
"grouped_alert_event_counts",
_RUNTIME_GROUPED_ALERT_EVENT_COUNTS_SQL,
)
log_controlled_writeback_consumer = (
await _load_log_controlled_writeback_consumer_readback(
project_id=project_id,
)
)
except Exception as exc:
logger.warning(
"ai_agent_autonomous_runtime_receipt_readback_failed",
project_id=project_id,
error_type=type(exc).__name__,
)
log_controlled_writeback_consumer = (
await _load_log_controlled_writeback_consumer_readback(
project_id=project_id,
)
)
return build_runtime_receipt_readback_from_rows(
project_id=project_id,
lookback_hours=params["lookback_hours"],
db_read_status="unavailable",
error_type=type(exc).__name__,
log_controlled_writeback_consumer=log_controlled_writeback_consumer,
)
log_controlled_writeback_consumer = (
await _load_log_controlled_writeback_consumer_readback(
project_id=project_id,
)
)
return build_runtime_receipt_readback_from_rows(
project_id=project_id,
lookback_hours=params["lookback_hours"],

View File

@@ -1,3 +1,6 @@
import pytest
from src.services import ai_agent_autonomous_runtime_control as runtime_control_module
from src.services.ai_agent_autonomous_runtime_control import (
_RUNTIME_ALERT_OPERATION_COUNTS_SQL,
_RUNTIME_ALERTMANAGER_EVENT_COUNTS_SQL,
@@ -10,6 +13,19 @@ from src.services.ai_agent_autonomous_runtime_control import (
)
class _FailingRuntimeDb:
async def execute(self, *_args, **_kwargs):
raise RuntimeError("runtime trace query unavailable")
class _FailingRuntimeDbContext:
async def __aenter__(self) -> _FailingRuntimeDb:
return _FailingRuntimeDb()
async def __aexit__(self, _exc_type, _exc, _tb) -> bool:
return False
def _assert_log_controlled_writeback_executor(payload: dict):
assert payload["schema_version"] == "ai_agent_log_controlled_writeback_executor_readback_v1"
assert payload["priority"] == "P1-LOG-KM-RAG-MCP-PLAYBOOK"
@@ -190,6 +206,62 @@ def _assert_log_controlled_writeback_consumer(payload: dict):
assert payload["operation_boundaries"]["github_api_used"] is False
@pytest.mark.asyncio
async def test_live_runtime_receipt_keeps_consumer_readback_when_trace_query_fails(
monkeypatch,
):
async def _fake_consumer_readback(*, project_id: str):
assert project_id == "awoooi"
return _log_controlled_writeback_consumer_readback()
monkeypatch.setattr(
runtime_control_module,
"get_db_context",
lambda project_id: _FailingRuntimeDbContext(),
)
monkeypatch.setattr(
runtime_control_module,
"_load_log_controlled_writeback_consumer_readback",
_fake_consumer_readback,
)
readback = (
await runtime_control_module.load_ai_agent_autonomous_runtime_receipt_readback()
)
assert readback["db_read_status"] == "unavailable"
assert readback["error"]["type"] == "RuntimeError"
_assert_log_controlled_writeback_consumer(
readback["log_controlled_writeback_consumer"]
)
def test_runtime_control_rollups_project_consumer_dispatch_ledger_if_trace_missing():
payload = build_ai_agent_autonomous_runtime_control()
readback = build_runtime_receipt_readback_from_rows(
project_id="awoooi",
db_read_status="unavailable",
log_controlled_writeback_consumer=_log_controlled_writeback_consumer_readback(),
)
runtime_control_module._attach_runtime_receipt_readback(payload, readback)
assert payload["rollups"]["live_log_controlled_writeback_dispatch_count"] == 6
assert (
payload["rollups"][
"live_log_controlled_writeback_consumer_dispatch_ledger_count"
]
== 6
)
assert payload["rollups"]["live_log_controlled_writeback_consumer_binding_count"] == 6
assert (
payload["rollups"]["live_log_controlled_writeback_consumer_ready_binding_count"]
== 6
)
assert payload["rollups"]["live_log_controlled_writeback_consumer_ready_count"] == 1
assert payload["rollups"]["live_log_controlled_writeback_consumer_blocker_count"] == 0
def test_runtime_receipt_auxiliary_sql_keeps_source_family_counts_schema_safe():
from src.services.ai_agent_autonomous_runtime_control import (
_RUNTIME_OPERATION_COUNTS_SQL,