hansjone/netx
1
0
Fork 0
mirror of https://github.com/hansjone/netx.git synced 2026-10-14 22:20:50 +08:00
Code Issues Projects Releases 1 Packages Wiki Activity Actions 4

feat(cutover): pair metric collection rounds and isolate manual routes

Browse source
This commit is contained in:
oliver 2026-10-11 08:46:46 +08:00
parent 32d7d969f6
commit d3618deff6
14 changed files with 809 additions and 154 deletions
Download patch file Download diff file Expand all files Collapse all files

25
netx_api/biz_state_scheduler.py
View file

@ -110,12 +110,21 @@ def _sync_cutover_hf_windows() -> None:
projects = db.query(BizMigrationProject).all()
for proj in projects:
want = _hf_window_status(proj)
tids = _all_hf_task_ids(proj, "old") + _all_hf_task_ids(proj, "new")
if tids and any((t := db.get(BizStateTask, tid)) and not t.collect_group_id for tid in tids):
try:
from .biz_migration.service import ensure_highfreq
ensure_highfreq(db, proj.id, collect_now=False)
except Exception:
_log.exception("upgrade cutover collection pairs failed project=%s", proj.id)
window_status = _hf_window_status(proj)
tids = _all_hf_task_ids(proj, "old") + _all_hf_task_ids(proj, "new")
for tid in tids:
task = db.get(BizStateTask, tid)
if not task:
continue
want = "paused" if task.collect_manual_only else window_status
purpose = str(getattr(task, "purpose", None) or "").strip()
if purpose and purpose != PURPOSE_CUTOVER_HF:
continue
@ -149,12 +158,26 @@ def _enqueue_due_tasks() -> int:
.filter(
BizStateTask.status == "running",
BizStateTask.collect_running.is_(False),
BizStateTask.collect_manual_only.is_(False),
)
.all()
)
due_ids: list[str] = []
seen_groups: set[str] = set()
now = _utcnow()
for task in tasks:
if task.collect_group_id:
if task.collect_group_id in seen_groups:
continue
seen_groups.add(task.collect_group_id)
peers = db.query(BizStateTask).filter(BizStateTask.collect_group_id == task.collect_group_id).all()
if len(peers) != 2 or any(p.collect_running or p.status != "running" or p.collect_manual_only for p in peers):
continue
anchors = [p.last_collect_started_at or p.last_collect_ended_at for p in peers]
anchor = max((a for a in anchors if a), default=None)
if anchor is None or (now - anchor).total_seconds() >= max(p.interval_sec or 60 for p in peers):
due_ids.append(task.id)
continue
interval = max(60, int(task.interval_sec or 300))
# Prefer start-based interval to avoid drift when collect duration
# approaches the interval (end-based: 40s collect + 60s → 100s cycle).