Add config-sync cycle retention alongside config history keep.

Mirror LLDP job retention so finished sync cycles are pruned by policy while keeping per-NE config version history separate.

Co-authored-by: Cursor <cursoragent@cursor.com>

netx_api/config_sync_recovery.py

@@ -8,7 +8,14 @@ from datetime import datetime
 from sqlalchemy.orm import Session
 
 from .config_sync_runner import dispatch_cycle
-from .config_sync_service import finalize_cycle, sync_cycle_progress
+from .config_sync_service import (
+    DEFAULT_CYCLE_KEEP,
+    _cycle_keep_value,
+    ensure_policy,
+    finalize_cycle,
+    prune_config_sync_cycles,
+    sync_cycle_progress,
+)
 from .models import ConfigSyncCycle, ConfigSyncTask
 
 _log = logging.getLogger("netx.config_sync.recovery")
@@ -57,7 +64,18 @@ def recover_config_sync_on_startup(db: Session) -> int:
         .all()
     )
     cycles = sorted(cycles, key=lambda c: c.created_at or datetime.min)
+    def _prune() -> None:
+        try:
+            keep = _cycle_keep_value(ensure_policy(db))
+        except Exception:
+            keep = DEFAULT_CYCLE_KEEP
+        try:
+            prune_config_sync_cycles(db, keep=keep)
+        except Exception:
+            _log.exception("config_sync recovery prune failed")
+
     if not cycles:
+        _prune()
         return 0
 
     # Single-flight hygiene: keep newest, close older interrupted cycles.
@@ -69,6 +87,7 @@ def recover_config_sync_on_startup(db: Session) -> int:
             primary.id,
         )
         _close_cycle(db, stale, reason="superseded_active_cycle")
+    _prune()
 
     cycle_id = str(primary.id)
     orphans = (

netx_api/config_sync_schemas.py

@@ -20,6 +20,7 @@ class ConfigSyncPolicyOut(BaseModel):
     scope_mode: str
     selected_targets: list[ConfigSyncTargetRef] = Field(default_factory=list)
     history_keep: int
+    cycle_keep: int = 30
     updated_at: datetime | None = None
 
 
@@ -30,6 +31,7 @@ class ConfigSyncPolicyUpdate(BaseModel):
     scope_mode: Literal["all", "selected"] | None = None
     selected_targets: list[ConfigSyncTargetRef] | None = None
     history_keep: int | None = Field(default=None, ge=0, le=30)
+    cycle_keep: int | None = Field(default=None, ge=0, le=200)
 
 
 class ConfigSyncCycleCreate(BaseModel):

netx_api/config_sync_service.py

@@ -47,6 +47,9 @@ def _utcnow() -> datetime:
     return datetime.utcnow()
 
 
+DEFAULT_CYCLE_KEEP = 30
+
+
 def ensure_policy(db: Session) -> ConfigSyncPolicy:
     row = db.get(ConfigSyncPolicy, POLICY_ID)
     if row is None:
@@ -57,6 +60,35 @@ def ensure_policy(db: Session) -> ConfigSyncPolicy:
     return row
 
 
+def prune_config_sync_cycles(db: Session, *, keep: int = DEFAULT_CYCLE_KEEP) -> int:
+    """Delete finished cycles beyond ``keep`` (newest kept). Active cycles always retained."""
+    keep = max(0, min(200, int(keep)))
+    finished = (
+        db.query(ConfigSyncCycle)
+        .filter(ConfigSyncCycle.status.in_(("success", "fail", "cancelled")))
+        .order_by(ConfigSyncCycle.created_at.desc())
+        .all()
+    )
+    to_drop = finished if keep == 0 else finished[keep:]
+    if not to_drop:
+        return 0
+    dropped = 0
+    for cycle in to_drop:
+        cid = str(cycle.id)
+        db.query(ConfigSyncTask).filter(ConfigSyncTask.cycle_id == cid).delete(
+            synchronize_session=False
+        )
+        db.delete(cycle)
+        dropped += 1
+    if dropped:
+        db.commit()
+    return dropped
+
+
+def _cycle_keep_value(row: ConfigSyncPolicy) -> int:
+    return max(0, min(200, int(getattr(row, "cycle_keep", None) or DEFAULT_CYCLE_KEEP)))
+
+
 def _targets_from_json(raw: Any) -> list[ConfigSyncTargetRef]:
     items: list[ConfigSyncTargetRef] = []
     if not isinstance(raw, list):
@@ -80,6 +112,7 @@ def policy_to_out(row: ConfigSyncPolicy) -> ConfigSyncPolicyOut:
         scope_mode=str(row.scope_mode or "all"),
         selected_targets=_targets_from_json(row.selected_targets),
         history_keep=max(0, min(30, int(row.history_keep if row.history_keep is not None else 3))),
+        cycle_keep=_cycle_keep_value(row),
         updated_at=row.updated_at,
     )
 
@@ -107,9 +140,12 @@ def update_policy(db: Session, body: ConfigSyncPolicyUpdate) -> ConfigSyncPolicy
         ]
     if "history_keep" in data and data["history_keep"] is not None:
         row.history_keep = max(0, min(30, int(data["history_keep"])))
+    if "cycle_keep" in data and data["cycle_keep"] is not None:
+        row.cycle_keep = max(0, min(200, int(data["cycle_keep"])))
     row.updated_at = _utcnow()
     db.commit()
     db.refresh(row)
+    prune_config_sync_cycles(db, keep=_cycle_keep_value(row))
     return policy_to_out(row)
 
 
@@ -443,7 +479,12 @@ def stop_cycle(db: Session, cycle_id: str) -> ConfigSyncCycleOut:
         _release_pool(cycle_id)
     except Exception:
         pass
-    return cycle_to_out(row)
+    out = cycle_to_out(row)
+    try:
+        prune_config_sync_cycles(db, keep=_cycle_keep_value(ensure_policy(db)))
+    except Exception:
+        _log.exception("prune_config_sync_cycles after stop failed")
+    return out
 
 
 def dashboard(db: Session) -> ConfigSyncDashboardOut:
@@ -690,3 +731,7 @@ def finalize_cycle(db: Session, cycle_id: str) -> None:
         cycle.error_message = ""
     cycle.ended_at = _utcnow()
     db.commit()
+    try:
+        prune_config_sync_cycles(db, keep=_cycle_keep_value(ensure_policy(db)))
+    except Exception:
+        _log.exception("prune_config_sync_cycles after finish failed")

netx_api/main.py

@@ -829,6 +829,9 @@ def on_startup() -> None:
                 )
             except Exception:
                 pass
+            conn.exec_driver_sql(
+                "ALTER TABLE config_sync_policy ADD COLUMN IF NOT EXISTS cycle_keep INTEGER DEFAULT 30"
+            )
     except Exception:
         _schedule_log.exception("startup: auth/port_traffic/topology schema migration failed")
     _reset_runtime_pause_flags()

netx_api/models.py

@@ -693,6 +693,8 @@ class ConfigSyncPolicy(Base):
     scope_mode: Mapped[str] = mapped_column(String(32), default="all")  # all | selected
     selected_targets: Mapped[list] = mapped_column(_JsonType, default=list)
     history_keep: Mapped[int] = mapped_column(Integer, default=3)
+    # Finished sync cycles to retain (newest kept); active cycles always kept.
+    cycle_keep: Mapped[int] = mapped_column(Integer, default=30)
     updated_at: Mapped[datetime] = mapped_column(DateTime, default=datetime.utcnow)