Home charts execution time per node type on a log scale, from a /observability/node-types rollup

This commit is contained in:
2026-10-15 16:57:07 +02:00
parent fa8b178866
commit 56fa34c2c9
9 changed files with 360 additions and 3 deletions
@@ -86,6 +86,17 @@ class FlowRollup(BaseModel):
last_error_ts: float | None = None
class NodeTypeTiming(BaseModel):
"""How long one kind of node takes to run, across every node of it."""
type: str
#: How many published nodes of this type ran inside the window.
nodes: int
executions: int
avg_ms: float
max_ms: float
class RunRow(BaseModel):
id: str
flow: str
@@ -428,6 +439,58 @@ def read_flow_rollups(
]
@router.get("/node-types", response_model=list[NodeTypeTiming])
def read_node_type_timings(
session: SessionDep,
controller: FlowControllerDep,
hours: Annotated[int, Query(ge=0, le=8760)] = 24,
) -> Any:
"""Execution time per node type, slowest first, to point at the hotspots.
The rollups are kept per node, so the type is joined on from the published
flows. A node deleted since it ran has no type left to count under.
"""
hours = _window_hours(hours)
types = {
f"{flow.name}.{node.id}": node.type
for flow in controller.store.read_all()
for node in flow.nodes
}
statement = (
sa_select(
col(MetricBucket.node),
func.sum(col(MetricBucket.executions)).label("executions"),
func.sum(col(MetricBucket.duration_sum_ms)).label("duration_sum_ms"),
func.max(col(MetricBucket.duration_max_ms)).label("max_ms"),
)
.where(col(MetricBucket.bucket) >= _since(hours))
.group_by(col(MetricBucket.node))
)
totals: dict[str, dict[str, float]] = {}
for row in session.execute(statement):
kind = types.get(row.node)
if kind is None or not row.executions:
continue
entry = totals.setdefault(
kind, {"nodes": 0, "executions": 0, "duration_sum_ms": 0.0, "max_ms": 0.0}
)
entry["nodes"] += 1
entry["executions"] += row.executions
entry["duration_sum_ms"] += row.duration_sum_ms
entry["max_ms"] = max(entry["max_ms"], row.max_ms)
timings = [
NodeTypeTiming(
type=kind,
nodes=int(entry["nodes"]),
executions=int(entry["executions"]),
avg_ms=round(entry["duration_sum_ms"] / entry["executions"], 3),
max_ms=round(entry["max_ms"], 3),
)
for kind, entry in totals.items()
]
return sorted(timings, key=lambda timing: timing.avg_ms, reverse=True)
@router.get("/runs", response_model=RunPage)
def read_runs(
session: SessionDep,