perl_d
5dc939a5ab
CI / lint (push) Skipped
CI / test (3.11) (push) Skipped
CI / test (3.12) (push) Skipped
CI / test (3.13) (push) Skipped
CI / lint (pull_request) Failing after 28s
CI / test (3.11) (pull_request) Skipped
CI / test (3.12) (pull_request) Skipped
CI / test (3.13) (pull_request) Skipped
127 lines
3.9 KiB
Python
127 lines
3.9 KiB
Python
from __future__ import annotations
|
|
|
|
from collections.abc import Callable
|
|
from dataclasses import dataclass
|
|
from typing import TypeAlias
|
|
|
|
from aarecommon.errors import exception_handler
|
|
from aarecommon.errors.exception_handler import (
|
|
AareException,
|
|
LoopCenteringFailed,
|
|
SmargonException,
|
|
TellException,
|
|
TransformationInvalidException,
|
|
)
|
|
|
|
ExceptionType: TypeAlias = type[AareException]
|
|
|
|
|
|
@dataclass(frozen=True)
|
|
class WatcherTrip:
|
|
name: str
|
|
threshold: int
|
|
streak: int
|
|
exception_class: str
|
|
|
|
def message(self) -> str:
|
|
return (
|
|
f"{self.streak} consecutive {self.name} errors across samples "
|
|
f"({self.exception_class}) — automation halted"
|
|
)
|
|
|
|
|
|
class RecurrenceWatcher:
|
|
def __init__(self, *, name: str, observes: ExceptionType, threshold: int):
|
|
self.name = name
|
|
self._observes = observes
|
|
self.threshold = max(1, int(threshold))
|
|
self._streak = 0
|
|
|
|
@property
|
|
def streak(self) -> int:
|
|
return self._streak
|
|
|
|
def reset(self) -> None:
|
|
self._streak = 0
|
|
|
|
def observe(self, exception_class: type | None) -> bool:
|
|
if exception_class is None:
|
|
self._streak = 0
|
|
return False
|
|
if not isinstance(exception_class, type) or not issubclass(exception_class, self._observes):
|
|
return False
|
|
self._streak += 1
|
|
return self._streak >= self.threshold
|
|
|
|
def maybe_trip(self, exception_class: type | None) -> WatcherTrip | None:
|
|
if not self.observe(exception_class):
|
|
return None
|
|
class_name = exception_class.__name__ if isinstance(exception_class, type) else "Unknown"
|
|
return WatcherTrip(
|
|
name=self.name,
|
|
threshold=self.threshold,
|
|
streak=self._streak,
|
|
exception_class=class_name,
|
|
)
|
|
|
|
|
|
DEFAULT_WATCHERS: tuple[tuple[str, ExceptionType, int], ...] = (
|
|
("tell", TellException, 5),
|
|
("alc", LoopCenteringFailed, 3),
|
|
("transformation", TransformationInvalidException, 3),
|
|
("smargon", SmargonException, 3),
|
|
)
|
|
|
|
|
|
def create_default_watchers(overrides: dict[str, int] | None = None) -> list[RecurrenceWatcher]:
|
|
thresholds = dict(overrides or {})
|
|
return [
|
|
RecurrenceWatcher(name=name, observes=observes, threshold=thresholds.get(name, threshold))
|
|
for name, observes, threshold in DEFAULT_WATCHERS
|
|
]
|
|
|
|
|
|
def redis_key_to_env_var(key: str) -> str:
|
|
"""Adapter from canonical Redis key shape to an env-var-safe name.
|
|
|
|
``aare:watchers:{bl}:{watcher}:threshold`` →
|
|
``AARE_WATCHERS_{BL}_{WATCHER}_THRESHOLD``.
|
|
|
|
Used by the GUI today since it has no Redis client; the loader's
|
|
``get_value`` argument keeps the Redis key shape authoritative so the
|
|
DAQ side can later swap in ``cfg.get`` without churn (plan §7)."""
|
|
return key.replace(":", "_").upper()
|
|
|
|
|
|
def load_watcher_threshold_overrides(
|
|
*, beamline: str, watcher_names: list[str], get_value: Callable[[str], str | bytes | None]
|
|
) -> dict[str, int]:
|
|
overrides: dict[str, int] = {}
|
|
for watcher_name in watcher_names:
|
|
key = f"aare:watchers:{beamline}:{watcher_name}:threshold"
|
|
raw_value = get_value(key)
|
|
if raw_value is None:
|
|
continue
|
|
if isinstance(raw_value, bytes):
|
|
raw_value = raw_value.decode("utf-8", errors="ignore")
|
|
try:
|
|
parsed = int(str(raw_value).strip())
|
|
except (TypeError, ValueError):
|
|
continue
|
|
if parsed > 0:
|
|
overrides[watcher_name] = parsed
|
|
return overrides
|
|
|
|
|
|
_EXCEPTION_CLASS_BY_NAME: dict[str, type] = {
|
|
name: klass
|
|
for name, klass in vars(exception_handler).items()
|
|
if isinstance(klass, type) and issubclass(klass, AareException)
|
|
}
|
|
|
|
|
|
def resolve_exception_class(exception_class_name: str | None) -> type | None:
|
|
if not exception_class_name:
|
|
return None
|
|
return _EXCEPTION_CLASS_BY_NAME.get(exception_class_name)
|