trex/trex-core
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: persist uncertain cancel outcomes

Browse Source
This commit is contained in:
jordan
2026-07-29 17:17:14 +09:00
parent a656350a35
commit 3dd8eefc92
21 changed files with 416 additions and 76 deletions
Download Patch File Download Diff File Expand all files Collapse all files
trex_core/adapters/broker/kis.py
+26 -5
View File
@@ -24,6 +24,8 @@ import httpx
from trex_core.application.broker_port import (
BalanceItem,
BrokerBalanceError,
CancelOutcome,
CancelState,
FillStatus,
OrderAck,
OrderRequest,
@@ -756,7 +758,7 @@ class KISBroker:
def cancel_order(
self, broker_order_id: str, *, side: str, symbol: str, qty: int, venue: str = "KRX"
) -> bool:
) -> CancelOutcome:
"""Cancel a previously submitted order (REQ-118).
venue: exchange venue of the original order. Must match the EXCG_ID_DVSN_CD
@@ -764,7 +766,10 @@ class KISBroker:
Real-account NXT cancel (EXCG=NXT) confirmed rt_cd=0 on 2026-07-06.
"""
if broker_order_id.startswith(_SUBMISSION_UNKNOWN_PREFIX):
return False
return CancelOutcome(
CancelState.UNKNOWN,
"submission_outcome_unknown",
)
cano, product = self._account()
parts = broker_order_id.split(":")
orgno, odno = (parts[0], parts[1]) if len(parts) == 2 else ("", parts[-1])
@@ -789,6 +794,22 @@ class KISBroker:
json=body,
timeout=self._timeout,
)
return str(resp.json().get("rt_cd")) == "0"
except (httpx.HTTPError, ValueError):
return False
except httpx.HTTPError:
return CancelOutcome(CancelState.UNKNOWN, "transport_failed")
if getattr(resp, "status_code", 200) >= 500:
return CancelOutcome(CancelState.UNKNOWN, "broker_unavailable")
try:
payload = resp.json()
except ValueError:
return CancelOutcome(CancelState.UNKNOWN, "malformed_response")
rt_cd = payload.get("rt_cd")
if str(rt_cd) == "0":
return CancelOutcome(CancelState.REQUESTED)
if rt_cd is None:
if getattr(resp, "status_code", 200) >= 400:
return CancelOutcome(CancelState.REJECTED, "http_rejected")
return CancelOutcome(CancelState.UNKNOWN, "missing_result_code")
return CancelOutcome(
CancelState.REJECTED,
str(payload.get("msg_cd") or "broker_rejected"),
)
trex_core/adapters/broker/ownership.py
+2 -1
View File
@@ -16,6 +16,7 @@ from threading import Lock
from trex_core.application.broker_port import (
BalanceItem,
BrokerBalanceError,
CancelOutcome,
FillStatus,
OrderAck,
OrderRequest,
@@ -146,7 +147,7 @@ class OwnershipAwareBroker:
symbol: str,
qty: int,
venue: str = "KRX",
) -> bool:
) -> CancelOutcome:
return self._delegate.cancel_order(
broker_order_id,
side=side,
trex_core/adapters/broker/paper.py
+13 -3
View File
@@ -14,6 +14,8 @@ from typing import Optional
from trex_core.adapters.broker.seq_store import SeqStore
from trex_core.application.broker_port import (
BalanceItem,
CancelOutcome,
CancelState,
FillStatus,
OrderAck,
OrderRequest,
@@ -105,16 +107,24 @@ class PaperBroker:
)
return FillStatus(state="partially_filled", filled_qty=o.filled, avg_price=o.price)
def cancel_order(self, broker_order_id: str, *, side: str, symbol: str, qty: int, venue: str = "KRX") -> bool:
def cancel_order(
self,
broker_order_id: str,
*,
side: str,
symbol: str,
qty: int,
venue: str = "KRX",
) -> CancelOutcome:
"""Cancel a paper order. venue is accepted for BrokerPort protocol conformance
(REQ-118) but ignored — paper has no exchange concept. Existing behavior preserved.
"""
with self._lock:
o = self._orders.get(broker_order_id)
if o is None or o.filled >= o.qty:
return False
return CancelOutcome(CancelState.REJECTED, "order_not_cancellable")
o.cancelled = True
return True
return CancelOutcome(CancelState.REQUESTED)
def get_balance(self) -> list[BalanceItem]:
"""Deterministic per-symbol fill aggregate (REQ-031 AC-3).
trex_core/adapters/persistence/postgresql.py
+32 -4
View File
@@ -37,7 +37,8 @@ _SLOT_COLUMNS = (
"order_id, strategy_id, broker_order_id, portfolio_id, "
"sell_strategy_id, forced_close_at, stop_loss_price, take_profit_price, "
"created_at, updated_at, instance_key, retired_at, "
"timeout_sec, slippage_pct, occupancy_seq, venue"
"timeout_sec, slippage_pct, occupancy_seq, venue, "
"cancel_state, cancel_requested_at"
)
@@ -216,7 +217,8 @@ class PostgreSQLSlotRepository:
"broker_order_id=%s, portfolio_id=%s, sell_strategy_id=%s, "
"forced_close_at=%s, stop_loss_price=%s, take_profit_price=%s, "
"instance_key=%s, retired_at=%s, timeout_sec=%s, slippage_pct=%s, "
"occupancy_seq=%s, venue=%s, updated_at=%s WHERE slot_id=%s",
"occupancy_seq=%s, venue=%s, cancel_state=%s, "
"cancel_requested_at=%s, updated_at=%s WHERE slot_id=%s",
(
slot.status.value,
slot.symbol,
@@ -237,6 +239,8 @@ class PostgreSQLSlotRepository:
slot.slippage_pct,
slot.occupancy_seq,
slot.venue,
slot.cancel_state,
slot.cancel_requested_at,
slot.updated_at or datetime.now(tz=timezone.utc),
slot.slot_id,
),
@@ -332,6 +336,26 @@ class PostgreSQLSlotRepository:
conn.commit()
return ok
def mark_cancel_outcome(
self,
slot_id: UUID,
broker_order_id: str,
state: str,
) -> bool:
now = datetime.now(tz=timezone.utc)
with self._psycopg.connect(self._dsn, autocommit=False) as conn:
with conn.cursor() as cur:
cur.execute(
"UPDATE daily_slot SET cancel_state=%s, "
"cancel_requested_at=%s, updated_at=%s "
"WHERE slot_id=%s AND broker_order_id=%s "
"AND status IN ('pending_buy','pending_sell')",
(state, now, now, slot_id, broker_order_id),
)
ok = cur.rowcount == 1
conn.commit()
return ok
def list_dispatched_pending(self, instance_key: str = "default") -> list[Slot]:
with self._psycopg.connect(self._dsn) as conn:
with conn.cursor() as cur:
@@ -347,7 +371,8 @@ class PostgreSQLSlotRepository:
with self._psycopg.connect(self._dsn, autocommit=False) as conn:
with conn.cursor() as cur:
cur.execute(
"UPDATE daily_slot SET broker_order_id=NULL, updated_at=%s WHERE slot_id=%s",
"UPDATE daily_slot SET broker_order_id=NULL, cancel_state=NULL, "
"cancel_requested_at=NULL, updated_at=%s WHERE slot_id=%s",
(datetime.now(tz=timezone.utc), slot_id),
)
conn.commit()
@@ -374,7 +399,8 @@ class PostgreSQLSlotRepository:
# forced_close_at(11) stop_loss_price(12) take_profit_price(13)
# created_at(14) updated_at(15) instance_key(16) retired_at(17)
# timeout_sec(18) slippage_pct(19) [dcy-022/REQ-068]
# occupancy_seq(20) [dcy-028/REQ-081], venue(21)
# occupancy_seq(20) [dcy-028/REQ-081], venue(21),
# cancel_state(22), cancel_requested_at(23)
return Slot(
slot_id=row[0] if isinstance(row[0], UUID) else UUID(str(row[0])),
status=SlotState(row[1]),
@@ -398,6 +424,8 @@ class PostgreSQLSlotRepository:
slippage_pct=f(row[19]),
occupancy_seq=int(row[20]) if len(row) > 20 and row[20] is not None else 0,
venue=str(row[21]) if len(row) > 21 and row[21] is not None else "KRX",
cancel_state=str(row[22]) if len(row) > 22 and row[22] is not None else None,
cancel_requested_at=row[23] if len(row) > 23 else None,
)
trex_core/adapters/persistence/sql/020_persist_cancel_request_outcome.sql
+18
View File
@@ -0,0 +1,18 @@
-- dcy-041 / GitHub #44: cancellation request acceptance is distinct from
-- final order cancellation. Persist the immediate result to prevent blind
-- automatic retries across execution ticks and daemon restarts.
ALTER TABLE daily_slot
ADD COLUMN IF NOT EXISTS cancel_state TEXT;
ALTER TABLE daily_slot
ADD COLUMN IF NOT EXISTS cancel_requested_at TIMESTAMPTZ;
ALTER TABLE daily_slot
DROP CONSTRAINT IF EXISTS daily_slot_cancel_state_check;
ALTER TABLE daily_slot
ADD CONSTRAINT daily_slot_cancel_state_check
CHECK (
cancel_state IS NULL
OR cancel_state IN ('requested', 'rejected', 'unknown')
);
trex_core/application/broker_port.py
+38 -2
View File
@@ -10,6 +10,7 @@ from __future__ import annotations
import os
from dataclasses import dataclass, field
from decimal import Decimal
from enum import Enum
from typing import Any, Optional, Protocol
@@ -29,6 +30,30 @@ class BrokerBalanceError(RuntimeError):
super().__init__(f"broker balance unavailable ({code})")
class CancelState(str, Enum):
"""Immediate result of a cancel request, not the final order state."""
REQUESTED = "requested"
REJECTED = "rejected"
UNKNOWN = "unknown"
@dataclass(frozen=True)
class CancelOutcome:
"""Sanitized cancel request outcome.
REQUESTED only means that the broker accepted the cancellation request.
Final cancelled/filled state is established exclusively by poll_fill().
"""
state: CancelState
reason: Optional[str] = None
@property
def is_requested(self) -> bool:
return self.state is CancelState.REQUESTED
@dataclass(frozen=True)
class OrderRequest:
slot_id: str # originating slot (audit/correlation)
@@ -114,12 +139,23 @@ class BrokerPort(Protocol):
"""Observe current fill status for a previously accepted order."""
...
def cancel_order(self, broker_order_id: str, *, side: str, symbol: str, qty: int, venue: str = "KRX") -> bool:
"""Best-effort cancel. Returns True if the broker accepted the cancel.
def cancel_order(
self,
broker_order_id: str,
*,
side: str,
symbol: str,
qty: int,
venue: str = "KRX",
) -> CancelOutcome:
"""Request cancellation and classify the immediate broker outcome.
venue: exchange venue of the original order (REQ-118). Cancel must use
the same EXCG_ID_DVSN_CD as the original submit. Default "KRX" preserves
backward compatibility for existing callers (orb/tsl).
REQUESTED is not a final cancellation. poll_fill() remains the only
authority for cancelled/filled terminal state.
"""
...
trex_core/application/execution.py
+83 -28
View File
@@ -29,7 +29,12 @@ from datetime import datetime, timezone
from typing import Callable, Optional
from trex_core.application.alarm_emitter import AlarmEmitter
from trex_core.application.broker_port import BrokerPort, OrderRequest
from trex_core.application.broker_port import (
BrokerPort,
CancelOutcome,
CancelState,
OrderRequest,
)
from trex_core.application.ledger_outbox import LedgerOutbox
from trex_core.application.ledger_port import LedgerTrade
from trex_core.domain.pricing import ceil_to_tick, floor_to_tick
@@ -546,12 +551,28 @@ class ExecutionService:
"""
if str(slot.broker_order_id or "").startswith("submission-unknown:"):
return
if slot.cancel_state is not None:
return
if slot.timeout_sec is None or slot.updated_at is None:
return
age = (now - slot.updated_at).total_seconds()
if age < slot.timeout_sec:
return
qty = int(slot.qty or 0)
# Write UNKNOWN before external I/O. If the process dies after KIS
# receives the request but before the response is persisted, restart
# polls the order instead of issuing a blind second cancellation.
if not self._repo.mark_cancel_outcome(
slot.slot_id,
str(slot.broker_order_id),
CancelState.UNKNOWN.value,
):
_log.error(
"cancel attempt reservation conflict: slot=%s broker_order_id=%s",
slot.slot_id,
slot.broker_order_id,
)
return
try:
cancel_kwargs = {
"side": side,
@@ -560,25 +581,40 @@ class ExecutionService:
}
if slot.venue != "KRX":
cancel_kwargs["venue"] = slot.venue
accepted = self._broker.cancel_order(slot.broker_order_id, **cancel_kwargs)
outcome = self._broker.cancel_order(slot.broker_order_id, **cancel_kwargs)
except Exception: # noqa: BLE001 - adapter failure becomes durable unknown
outcome = CancelOutcome(CancelState.UNKNOWN, "adapter_exception")
_log.warning(
"timeout cancel outcome unknown: slot=%s broker_order_id=%s",
slot.slot_id,
slot.broker_order_id,
)
persisted = self._repo.mark_cancel_outcome(
slot.slot_id,
str(slot.broker_order_id),
outcome.state.value,
)
if not persisted:
_log.error(
"cancel outcome persistence conflict: slot=%s broker_order_id=%s "
"outcome=%s",
slot.slot_id,
slot.broker_order_id,
outcome.state.value,
)
return
if outcome.is_requested:
r.timeout_cancelled += 1
_log.warning(
"timeout cancel sent: slot=%s status=%s broker_order_id=%s "
"age_sec=%.0f timeout_sec=%d broker_accepted=%s",
slot.slot_id,
slot.status.value,
slot.broker_order_id,
age,
slot.timeout_sec,
accepted,
)
except Exception: # noqa: BLE001 - best-effort; do not block the loop
_log.warning(
"timeout cancel failed (exception): slot=%s broker_order_id=%s "
"(will retry next tick if still pending)",
slot.slot_id,
slot.broker_order_id,
)
_log.warning(
"timeout cancel result: slot=%s status=%s broker_order_id=%s "
"age_sec=%.0f timeout_sec=%d outcome=%s",
slot.slot_id,
slot.status.value,
slot.broker_order_id,
age,
slot.timeout_sec,
outcome.state.value,
)
def _emit_event(self, event_type: str, slot, routing_strategy_id: Optional[str] = None) -> None:
if self._event_emitter is None:
@@ -655,9 +691,9 @@ class ExecutionService:
def cancel_pending_slot(self, slot_id) -> str:
"""Operator backstop: send cancel_order for a pending slot (dcy-022/REQ-071).
Returns "ok" (cancel sent), "not_pending" (slot is HOLDING/INACTIVE),
or "slot_not_found". Does NOT transition the slot — the execution loop's
next poll handles the rollback via the normal is_terminal_reject path.
Returns "ok" (request accepted), "cancel_rejected", "cancel_unknown",
"not_pending", "not_dispatched", or "slot_not_found". Does NOT
transition the slot — fill polling alone confirms cancelled/filled.
"""
from uuid import UUID
@@ -680,21 +716,40 @@ class ExecutionService:
}
if slot.venue != "KRX":
cancel_kwargs["venue"] = slot.venue
self._broker.cancel_order(slot.broker_order_id, **cancel_kwargs)
except Exception: # noqa: BLE001 - best-effort
# Durable unknown reservation closes the crash window between the
# outbound request and saving its immediate result.
if not self._repo.mark_cancel_outcome(
slot.slot_id,
str(slot.broker_order_id),
CancelState.UNKNOWN.value,
):
return "cancel_state_conflict"
outcome = self._broker.cancel_order(slot.broker_order_id, **cancel_kwargs)
except Exception: # noqa: BLE001 - adapter failure becomes durable unknown
outcome = CancelOutcome(CancelState.UNKNOWN, "adapter_exception")
_log.warning(
"operator cancel failed: slot=%s broker_order_id=%s",
"operator cancel outcome unknown: slot=%s broker_order_id=%s",
slot.slot_id,
slot.broker_order_id,
)
return "cancel_failed"
if not self._repo.mark_cancel_outcome(
slot.slot_id,
str(slot.broker_order_id),
outcome.state.value,
):
return "cancel_state_conflict"
_log.warning(
"operator cancel sent: slot=%s status=%s broker_order_id=%s",
"operator cancel result: slot=%s status=%s broker_order_id=%s outcome=%s",
slot.slot_id,
slot.status.value,
slot.broker_order_id,
outcome.state.value,
)
return "ok"
if outcome.state is CancelState.REQUESTED:
return "ok"
if outcome.state is CancelState.REJECTED:
return "cancel_rejected"
return "cancel_unknown"
def _rollback(self, slot, *, reason: str) -> None:
total = self._balance_map().get(slot.symbol or "", 0)
trex_core/domain/slot/in_memory_repository.py
+28 -1
View File
@@ -219,6 +219,29 @@ class InMemorySlotRepository:
)
return True
def mark_cancel_outcome(
self,
slot_id: UUID,
broker_order_id: str,
state: str,
) -> bool:
"""Persist one cancel attempt iff the same order is still pending."""
s = self._slots.get(slot_id)
if (
s is None
or s.broker_order_id != broker_order_id
or s.status not in self._PENDING
):
return False
now = datetime.now(tz=timezone.utc)
self._slots[slot_id] = replace(
s,
cancel_state=state,
cancel_requested_at=now,
updated_at=now,
)
return True
def list_dispatched_pending(self, instance_key: str = "default") -> list[Slot]:
"""Pending slots already dispatched (broker_order_id NOT NULL) — poll targets."""
return [
@@ -236,7 +259,11 @@ class InMemorySlotRepository:
s = self._slots.get(slot_id)
if s is not None:
self._slots[slot_id] = replace(
s, broker_order_id=None, updated_at=datetime.now(tz=timezone.utc)
s,
broker_order_id=None,
cancel_state=None,
cancel_requested_at=None,
updated_at=datetime.now(tz=timezone.utc),
)
# --- bulk counts ---
trex_core/domain/slot/models.py
+7
View File
@@ -59,6 +59,11 @@ class Slot:
# Order venue carried from the admitted intent through dispatch/cancel.
# It is reset to KRX when the permanent asset slot returns to INACTIVE.
venue: str = "KRX"
# Immediate cancel-request result. The slot remains pending until fill
# polling confirms cancelled/filled. Persisted to prevent blind retries
# across ticks and restarts.
cancel_state: Optional[str] = None
cancel_requested_at: Optional[datetime] = None
# Canonical column contract for `Slot` (dcy-001 OQ-2 / FR-5).
@@ -88,6 +93,8 @@ EXPECTED_SLOT_FIELDS: frozenset[str] = frozenset({
"slippage_pct", # dcy-022 / REQ-068 (BAT-032)
"occupancy_seq", # dcy-028 / REQ-081 (BAT-042)
"venue",
"cancel_state",
"cancel_requested_at",
})
trex_core/domain/slot/service.py
+4
View File
@@ -205,6 +205,8 @@ class SlotService:
timeout_sec=None,
slippage_pct=None,
venue="KRX",
cancel_state=None,
cancel_requested_at=None,
updated_at=datetime.now(tz=timezone.utc),
)
@@ -277,6 +279,8 @@ class SlotService:
slippage_pct=new_slippage_pct,
occupancy_seq=new_occupancy_seq,
venue=new_venue,
cancel_state=None,
cancel_requested_at=None,
updated_at=datetime.now(tz=timezone.utc),
)