feat: host heartbeat status and dashboard metrics (2.2)

Track agent.heartbeat and report.daily.ssh per host; show online/stale
on Hosts UI; dashboard counters and SAC_HEARTBEAT_STALE_MINUTES config.
This commit is contained in:
2026-05-27 13:36:17 +10:00
parent 9ba3c4b63b
commit 04301ab359
14 changed files with 246 additions and 4 deletions
+53
View File
@@ -0,0 +1,53 @@
"""Статус агентов по событиям agent.heartbeat / report.daily.* в БД."""
from datetime import datetime, timezone
from sqlalchemy import func, select
from sqlalchemy.orm import Session
from app.models import Event, Host
HEARTBEAT_TYPE = "agent.heartbeat"
DAILY_REPORT_SSH = "report.daily.ssh"
def max_event_time_by_host(db: Session, event_type: str) -> dict[int, datetime]:
rows = db.execute(
select(Event.host_id, func.max(Event.received_at))
.where(Event.type == event_type)
.group_by(Event.host_id)
).all()
return {int(host_id): ts for host_id, ts in rows if ts is not None}
def agent_status(
last_heartbeat_at: datetime | None,
*,
stale_minutes: int,
now: datetime | None = None,
) -> str:
"""
online — heartbeat не старше порога;
stale — был хост, но heartbeat устарел или отсутствует;
unknown — heartbeat ещё не приходил.
"""
if last_heartbeat_at is None:
return "unknown"
ref = now or datetime.now(timezone.utc)
hb = last_heartbeat_at
if hb.tzinfo is None:
hb = hb.replace(tzinfo=timezone.utc)
age_sec = (ref - hb).total_seconds()
if age_sec > stale_minutes * 60:
return "stale"
return "online"
def count_stale_hosts(db: Session, stale_minutes: int) -> int:
hb_map = max_event_time_by_host(db, HEARTBEAT_TYPE)
host_ids = db.scalars(select(Host.id)).all()
return sum(
1
for host_id in host_ids
if agent_status(hb_map.get(host_id), stale_minutes=stale_minutes) == "stale"
)