Merge branch 'feature/metrics-history'

.gitignore

@@ -74,3 +74,6 @@ graphify-out/
 
 # Claude Code local preview config
 .claude/launch.json
+# Local-only Superpowers artifacts (specs, plans, SDD scratch) — never pushed
+docs/superpowers/
+.superpowers/

backend/app/api/proxmox/vms.py

@@ -21,6 +21,7 @@ from ...ipam_service import IPAMService
 from ._helpers import (check_vm_access, require_vm_access, _get_proxmox_client,
                         get_next_vmid, archive_and_delete_snapshots,
                         save_vm_instance, get_vm_instance, soft_delete_vm_instance)
+from ...services.metrics_history import query_instance_metrics, timeframe_to_range
 
 router = APIRouter()
 
@@ -1868,6 +1869,55 @@ def get_container_rrddata(
         raise HTTPException(status_code=500, detail=str(e))
 
 
+# ==================== Metrics History (DB-only) ====================
+
+def _resolve_metrics_window(timeframe, from_ts, to_ts):
+    from ...config import settings
+    now = int(time_lib.time())
+    if from_ts is not None and to_ts is not None:
+        f, t = int(from_ts), int(to_ts)
+    else:
+        f, t = timeframe_to_range(timeframe, now)
+    max_span = getattr(settings, "METRICS_RETENTION_DAYS", 30) * 86400
+    if t - f > max_span:
+        f = t - max_span
+    return f, t
+
+
+@router.get("/api/{server_id}/vm/{vmid}/metrics")
+def get_vm_metrics(
+    server_id: int,
+    vmid: int,
+    node: str = Query(None),
+    timeframe: str = Query("hour", regex="^(hour|day|week|month)$"),
+    from_ts: int = Query(None),
+    to_ts: int = Query(None),
+    nic: str = Query("all"),
+    db: Session = Depends(get_db),
+    current_user: User = Depends(PermissionChecker("vm:view")),
+):
+    """Retrieve VM metrics time-series from the local DB (no Proxmox call)."""
+    f, t = _resolve_metrics_window(timeframe, from_ts, to_ts)
+    return query_instance_metrics(db, server_id, vmid, f, t, nic)
+
+
+@router.get("/api/{server_id}/container/{vmid}/metrics")
+def get_container_metrics(
+    server_id: int,
+    vmid: int,
+    node: str = Query(None),
+    timeframe: str = Query("hour", regex="^(hour|day|week|month)$"),
+    from_ts: int = Query(None),
+    to_ts: int = Query(None),
+    nic: str = Query("all"),
+    db: Session = Depends(get_db),
+    current_user: User = Depends(PermissionChecker("vm:view")),
+):
+    """Retrieve container metrics time-series from the local DB (no Proxmox call)."""
+    f, t = _resolve_metrics_window(timeframe, from_ts, to_ts)
+    return query_instance_metrics(db, server_id, vmid, f, t, nic)
+
+
 # ==================== VNC Console ====================
 
 @router.get("/api/{server_id}/vm/{vmid}/vnc")
@@ -2850,6 +2900,7 @@ def set_vm_owner(
     if not instance:
         raise HTTPException(status_code=404, detail="VM not found in cache")
 
+    target_user = None
     if body.user_id is not None:
         target_user = db.query(User).filter(User.id == body.user_id, User.is_active == True).first()
         if not target_user:
@@ -2858,6 +2909,29 @@ def set_vm_owner(
     instance.owner_id = body.user_id
     db.commit()
     logger.info(f"Admin {current_user.username} set owner of VM {vmid} (server {server_id}) to user_id={body.user_id}")
+
+    # Push a real-time event so every connected client refreshes the owner
+    # column without a manual page reload.
+    try:
+        from ...websocket_manager import broadcast_event
+        owner_user = (
+            {
+                "username": target_user.username,
+                "email": target_user.email,
+                "full_name": target_user.full_name,
+            }
+            if target_user else None
+        )
+        broadcast_event(
+            "vm_owner_changed",
+            server_id=server_id,
+            vmid=vmid,
+            owner_id=body.user_id,
+            owner_user=owner_user,
+        )
+    except Exception as e:
+        logger.debug(f"Failed to broadcast vm_owner_changed: {e}")
+
     return {"ok": True, "owner_id": body.user_id}
 
 

backend/app/config.py

@@ -52,6 +52,10 @@ class Settings(BaseSettings):
     # once your servers' host keys are in known_hosts.
     SSH_REJECT_UNKNOWN_HOSTS: bool = Field(default=False, env="SSH_REJECT_UNKNOWN_HOSTS")
     
+    # Metrics history
+    METRICS_COLLECT_INTERVAL: int = Field(default=15, env="METRICS_COLLECT_INTERVAL")
+    METRICS_RETENTION_DAYS: int = Field(default=30, env="METRICS_RETENTION_DAYS")
+
     # CORS — comma-separated list of allowed origins; "*" is insecure in production
     CORS_ORIGINS: str = Field(default="*", env="CORS_ORIGINS")
 

backend/app/models/__init__.py

@@ -3,6 +3,7 @@ from .auth import Role, user_servers, User, ActiveSession, LoginAttempt, Blocked
 from .proxmox import OSTemplateGroup, OSTemplate, ProxmoxServer, VMInstance, VMSnapshotArchive
 from .ipam import IPAMNetwork, IPAMPool, IPAMAllocation, IPAMHistory
 from .tasks import TaskQueue, ProxmoxTask, DeployTask
+from .metrics import InstanceMetric, InstanceNicMetric
 from .settings import PanelSettings, SecuritySetting, Workspace, WorkspaceServer, WorkspaceUser
 from .misc import Notification, NotificationPreference, BackupJob, AuditLog
 
@@ -12,6 +13,7 @@ __all__ = [
     'OSTemplateGroup', 'OSTemplate', 'ProxmoxServer', 'VMInstance', 'VMSnapshotArchive',
     'IPAMNetwork', 'IPAMPool', 'IPAMAllocation', 'IPAMHistory',
     'TaskQueue', 'ProxmoxTask', 'DeployTask',
+    'InstanceMetric', 'InstanceNicMetric',
     'PanelSettings', 'SecuritySetting', 'Workspace', 'WorkspaceServer', 'WorkspaceUser',
     'Notification', 'NotificationPreference', 'BackupJob', 'AuditLog'
 ]

backend/app/models/metrics.py

@@ -0,0 +1,39 @@
+from sqlalchemy import Float
+from .base import Base, Column, Integer, BigInteger, String, DateTime, Index, func
+
+
+class InstanceMetric(Base):
+    __tablename__ = "instance_metric"
+
+    id = Column(BigInteger().with_variant(Integer, "sqlite"), primary_key=True, autoincrement=True)
+    server_id = Column(Integer, nullable=False)
+    vmid = Column(Integer, nullable=False)
+    vm_type = Column(String(10), nullable=True)
+    ts = Column(DateTime(timezone=True), nullable=False, server_default=func.now())
+    cpu = Column(Float, nullable=True)
+    mem = Column(BigInteger, nullable=True)
+    maxmem = Column(BigInteger, nullable=True)
+    disk_used = Column(BigInteger, nullable=True)
+    disk_total = Column(BigInteger, nullable=True)
+    netin_rate = Column(Float, nullable=True)
+    netout_rate = Column(Float, nullable=True)
+    diskread_rate = Column(Float, nullable=True)
+    diskwrite_rate = Column(Float, nullable=True)
+    iops_read = Column(Float, nullable=True)
+    iops_write = Column(Float, nullable=True)
+
+    __table_args__ = (Index("ix_instance_metric_lookup", "server_id", "vmid", "ts"),)
+
+
+class InstanceNicMetric(Base):
+    __tablename__ = "instance_nic_metric"
+
+    id = Column(BigInteger().with_variant(Integer, "sqlite"), primary_key=True, autoincrement=True)
+    server_id = Column(Integer, nullable=False)
+    vmid = Column(Integer, nullable=False)
+    ts = Column(DateTime(timezone=True), nullable=False, server_default=func.now())
+    dev = Column(String(64), nullable=False)
+    in_rate = Column(Float, nullable=True)
+    out_rate = Column(Float, nullable=True)
+
+    __table_args__ = (Index("ix_instance_nic_metric_lookup", "server_id", "vmid", "ts"),)

backend/app/proxmox/client.py

@@ -1310,6 +1310,29 @@ class ProxmoxClient(VmMixin, LxcMixin, ClusterMixin, StorageMixin, NetworkMixin,
                 logger.debug(f"Не удалось получить сетевые интерфейсы VM {vmid} (guest agent может быть не установлен): {e}")
                 return []
 
+        def get_vm_blockstats(self, node: str, vmid: int) -> str:
+            """Raw HMP 'info blockstats' text via the QEMU monitor (VM only).
+            Returns '' if unavailable."""
+            if not self.proxmox:
+                return ""
+            try:
+                res = self.proxmox.nodes(node).qemu(vmid).monitor.post(command="info blockstats")
+                return res if isinstance(res, str) else str(res or "")
+            except Exception as e:
+                logger.debug(f"blockstats unavailable for VM {vmid} on {node}: {e}")
+                return ""
+
+        def get_node_netstat(self, node: str) -> list:
+            """Per-tap/veth interface counters for the node ([] on failure)."""
+            if not self.proxmox:
+                return []
+            try:
+                res = self.proxmox.nodes(node).netstat.get()
+                return res if isinstance(res, list) else []
+            except Exception as e:
+                logger.debug(f"netstat unavailable for node {node}: {e}")
+                return []
+
         def get_container_interfaces(self, node: str, vmid: int) -> List[Dict]:
             """
             Получить сетевые интерфейсы контейнера

backend/app/services/metrics_broadcaster.py

@@ -606,4 +606,8 @@ def start_metrics_broadcaster(db_factory) -> None:
     asyncio.create_task(server_metrics_loop(db_factory), name="metrics:servers")
     asyncio.create_task(instance_metrics_loop(db_factory), name="metrics:instances")
     asyncio.create_task(instances_list_metrics_loop(db_factory), name="metrics:instances_list")
+    # Import inside function to avoid circular imports
+    from .metrics_collector import metrics_collector_loop, metrics_retention_loop
+    asyncio.create_task(metrics_collector_loop(db_factory), name="metrics:collector")
+    asyncio.create_task(metrics_retention_loop(db_factory), name="metrics:retention")
     logger.info("[metrics_broadcaster] Metrics broadcasting tasks started")

backend/app/services/metrics_collector.py

@@ -0,0 +1,148 @@
+"""Background collector that persists instance telemetry to our own tables."""
+from __future__ import annotations
+
+import asyncio
+import time
+from datetime import datetime, timezone
+
+from loguru import logger
+
+from .metrics_history import (
+    compute_rate, parse_blockstats, parse_netstat, aggregate_fsinfo,
+)
+
+
+def _status(client, inst_type, node, vmid):
+    if inst_type == "qemu":
+        return client.get_vm_status(node, vmid)
+    return client.get_container_status(node, vmid)
+
+
+def collect_once(db, items, clients, prev, _now=None):
+    """One collection tick. Mutates `prev` to carry counters between ticks."""
+    from ..models import InstanceMetric, InstanceNicMetric
+    now = _now if _now is not None else time.monotonic()
+    ts = datetime.now(tz=timezone.utc)
+
+    # netstat is one call per (server_id, node); cache per tick.
+    netstat_cache: dict[tuple, list] = {}
+    # Track which prev keys are touched this tick so stale entries can be evicted.
+    seen: set = set()
+
+    for it in items:
+        if it.get("status") != "running":
+            continue
+        sid, vmid = it["server_id"], it["vmid"]
+        inst_type, node = it.get("type"), it.get("node")
+        client = clients.get(sid)
+        if client is None or not node:
+            continue
+        try:
+            status = _status(client, inst_type, node, vmid)
+            if not status:
+                continue
+
+            # disk fill
+            if inst_type == "qemu":
+                used, total = aggregate_fsinfo(client.get_vm_fsinfo(node, vmid))
+            else:
+                used = int(status.get("disk", 0) or 0)
+                total = int(status.get("maxdisk", 0) or 0)
+
+            # cumulative counters → rates
+            key = f"{sid}:{vmid}"
+            cur = {
+                "netin": float(status.get("netin", 0) or 0),
+                "netout": float(status.get("netout", 0) or 0),
+                "diskread": float(status.get("diskread", 0) or 0),
+                "diskwrite": float(status.get("diskwrite", 0) or 0),
+            }
+            if inst_type == "qemu":
+                rd_ops, wr_ops = parse_blockstats(client.get_vm_blockstats(node, vmid))
+                cur["rd_ops"], cur["wr_ops"] = float(rd_ops), float(wr_ops)
+
+            pkey = prev.get(key)
+            dt = (now - pkey[0]) if pkey else 0
+            pc = pkey[1] if pkey else {}
+
+            def rate(name):
+                return compute_rate(pc.get(name), cur.get(name), dt)
+
+            db.add(InstanceMetric(
+                server_id=sid, vmid=vmid, vm_type=inst_type, ts=ts,
+                cpu=status.get("cpu"), mem=status.get("mem"), maxmem=status.get("maxmem"),
+                disk_used=used, disk_total=total,
+                netin_rate=rate("netin"), netout_rate=rate("netout"),
+                diskread_rate=rate("diskread"), diskwrite_rate=rate("diskwrite"),
+                iops_read=rate("rd_ops") if inst_type == "qemu" else None,
+                iops_write=rate("wr_ops") if inst_type == "qemu" else None,
+            ))
+            prev[key] = (now, cur)
+            seen.add(key)
+
+            # per-NIC
+            ck = (sid, node)
+            if ck not in netstat_cache:
+                netstat_cache[ck] = client.get_node_netstat(node)
+            per_dev = parse_netstat(netstat_cache[ck], vmid)
+            for dev, (din, dout) in per_dev.items():
+                nkey = f"{sid}:{vmid}:{dev}"
+                npkey = prev.get(nkey)
+                ndt = (now - npkey[0]) if npkey else 0
+                npc = npkey[1] if npkey else {}
+                db.add(InstanceNicMetric(
+                    server_id=sid, vmid=vmid, ts=ts, dev=dev,
+                    in_rate=compute_rate(npc.get("in"), float(din), ndt),
+                    out_rate=compute_rate(npc.get("out"), float(dout), ndt),
+                ))
+                prev[nkey] = (now, {"in": float(din), "out": float(dout)})
+                seen.add(nkey)
+        except Exception as e:
+            logger.debug(f"[metrics_collector] {sid}:{vmid} error: {e}")
+
+    # Evict stale keys from prev (VMs/NICs not seen this tick).
+    for k in list(prev):
+        if k not in seen:
+            del prev[k]
+
+
+async def metrics_collector_loop(db_factory) -> None:
+    from ..config import settings
+    from .metrics_broadcaster import _enumerate_instances_for_list
+    prev: dict = {}
+    while True:
+        start = time.monotonic()
+        try:
+            items, clients = await asyncio.to_thread(_enumerate_instances_for_list, db_factory)
+            db = db_factory()
+            try:
+                await asyncio.to_thread(collect_once, db, items, clients, prev)
+                db.commit()
+            finally:
+                db.close()
+        except asyncio.CancelledError:
+            break
+        except Exception as e:
+            logger.warning(f"[metrics_collector] tick error: {e}")
+        await asyncio.sleep(max(1.0, settings.METRICS_COLLECT_INTERVAL - (time.monotonic() - start)))
+
+
+async def metrics_retention_loop(db_factory) -> None:
+    from ..config import settings
+    from ..models import InstanceMetric, InstanceNicMetric
+    from datetime import timedelta
+    while True:
+        try:
+            cutoff = datetime.now(tz=timezone.utc) - timedelta(days=settings.METRICS_RETENTION_DAYS)
+            db = db_factory()
+            try:
+                db.query(InstanceMetric).filter(InstanceMetric.ts < cutoff).delete()
+                db.query(InstanceNicMetric).filter(InstanceNicMetric.ts < cutoff).delete()
+                db.commit()
+            finally:
+                db.close()
+        except asyncio.CancelledError:
+            break
+        except Exception as e:
+            logger.warning(f"[metrics_collector] retention error: {e}")
+        await asyncio.sleep(3600)

backend/app/services/metrics_history.py

@@ -0,0 +1,163 @@
+"""Pure helpers + read query for instance metric history."""
+from __future__ import annotations
+
+import re
+from datetime import timezone as _tz
+
+_RD_OPS = re.compile(r"rd_operations=(\d+)")
+_WR_OPS = re.compile(r"wr_operations=(\d+)")
+
+
+def compute_rate(prev, cur, dt):
+    """Per-second rate from cumulative counters. None if unknowable; 0 on reset."""
+    if prev is None or cur is None or not dt or dt <= 0:
+        return None
+    delta = float(cur) - float(prev)
+    if delta < 0:
+        return 0.0
+    return delta / dt
+
+
+def parse_blockstats(text: str):
+    """Sum rd_operations / wr_operations across all block devices in HMP output."""
+    rd = sum(int(m) for m in _RD_OPS.findall(text or ""))
+    wr = sum(int(m) for m in _WR_OPS.findall(text or ""))
+    return rd, wr
+
+
+def parse_netstat(rows, vmid: int):
+    """Map dev -> (in_bytes, out_bytes) for the given vmid from /nodes/{node}/netstat."""
+    out: dict[str, tuple[int, int]] = {}
+    for r in rows or []:
+        try:
+            if int(r.get("vmid")) != int(vmid):
+                continue
+        except (TypeError, ValueError):
+            continue
+        dev = r.get("dev")
+        if not dev:
+            continue
+        out[dev] = (int(float(r.get("in", 0) or 0)), int(float(r.get("out", 0) or 0)))
+    return out
+
+
+def aggregate_fsinfo(disks):
+    """Sum used/total bytes across guest filesystems."""
+    used = sum(int(d.get("used", 0) or 0) for d in disks or [])
+    total = sum(int(d.get("total", 0) or 0) for d in disks or [])
+    return used, total
+
+
+_TIMEFRAME_SECONDS = {
+    "hour": 3600,
+    "day": 86400,
+    "week": 7 * 86400,
+    "month": 30 * 86400,
+}
+
+
+def timeframe_to_range(timeframe, now_ts):
+    """Convert timeframe string to (from_ts, to_ts) tuple."""
+    span = _TIMEFRAME_SECONDS.get(timeframe, 3600)
+    return now_ts - span, now_ts
+
+
+def pick_bucket_seconds(from_ts, to_ts, target_points=150, floor=15):
+    """Calculate bucket size in seconds for time-series aggregation."""
+    span = max(1, to_ts - from_ts)
+    ideal = span // max(1, target_points)
+    return max(floor, int(ideal))
+
+
+def aggregate_series(rows, from_ts, bucket, fields):
+    """Aggregate time-series data into buckets, averaging numeric fields."""
+    if not rows:
+        return []
+    buckets: dict[int, dict] = {}
+    order: list[int] = []
+    for r in rows:
+        b = from_ts + ((int(r["time"]) - from_ts) // bucket) * bucket
+        if b not in buckets:
+            buckets[b] = {f: [] for f in fields}
+            order.append(b)
+        for f in fields:
+            v = r.get(f)
+            if v is not None:
+                buckets[b][f].append(float(v))
+    out = []
+    for b in order:
+        point = {"time": b}
+        for f in fields:
+            vals = buckets[b][f]
+            point[f] = (sum(vals) / len(vals)) if vals else None
+        out.append(point)
+    return out
+
+
+_AGG_FIELDS = ["cpu", "mem", "maxmem", "netin", "netout",
+               "diskread", "diskwrite", "diskpct", "iops_read", "iops_write"]
+
+
+def _to_ts(dt):
+    if dt.tzinfo is None:
+        dt = dt.replace(tzinfo=_tz.utc)
+    return int(dt.timestamp())
+
+
+def query_instance_metrics(db, server_id, vmid, from_ts, to_ts, nic="all"):
+    from ..models import InstanceMetric, InstanceNicMetric
+    from datetime import datetime
+
+    start = datetime.fromtimestamp(from_ts, tz=_tz.utc)
+    end = datetime.fromtimestamp(to_ts, tz=_tz.utc)
+
+    base = (db.query(InstanceMetric)
+              .filter(InstanceMetric.server_id == server_id,
+                      InstanceMetric.vmid == vmid,
+                      InstanceMetric.ts >= start, InstanceMetric.ts <= end)
+              .order_by(InstanceMetric.ts.asc()).all())
+
+    # Distinct NICs seen in the window (for the selector).
+    if nic == "all":
+        # Only fetch distinct dev names — avoids loading all NIC rows.
+        dev_rows = (db.query(InstanceNicMetric.dev)
+                      .filter(InstanceNicMetric.server_id == server_id,
+                              InstanceNicMetric.vmid == vmid,
+                              InstanceNicMetric.ts >= start, InstanceNicMetric.ts <= end)
+                      .distinct().all())
+        nics = sorted({r.dev for r in dev_rows})
+        nic_by_ts: dict[int, tuple] = {}
+    else:
+        # Load full NIC rows for the requested dev to build per-ts override map.
+        nic_rows = (db.query(InstanceNicMetric)
+                      .filter(InstanceNicMetric.server_id == server_id,
+                              InstanceNicMetric.vmid == vmid,
+                              InstanceNicMetric.ts >= start, InstanceNicMetric.ts <= end)
+                      .order_by(InstanceNicMetric.ts.asc()).all())
+        nics = sorted({r.dev for r in nic_rows})
+        nic_by_ts = {}
+        for r in nic_rows:
+            if r.dev == nic:
+                nic_by_ts[_to_ts(r.ts)] = (r.in_rate, r.out_rate)
+
+    rows = []
+    for m in base:
+        ts = _to_ts(m.ts)
+        netin, netout = m.netin_rate, m.netout_rate
+        if nic != "all":
+            ov = nic_by_ts.get(ts)
+            netin, netout = (ov if ov else (None, None))
+        diskpct = (m.disk_used / m.disk_total * 100.0) if (m.disk_total or 0) > 0 else None
+        rows.append({
+            "time": ts,
+            "cpu": (m.cpu * 100.0) if m.cpu is not None else None,
+            "mem": m.mem, "maxmem": m.maxmem,
+            "netin": netin, "netout": netout,
+            "diskread": m.diskread_rate, "diskwrite": m.diskwrite_rate,
+            "diskpct": diskpct,
+            "iops_read": m.iops_read, "iops_write": m.iops_write,
+        })
+
+    bucket = pick_bucket_seconds(from_ts, to_ts)
+    data = aggregate_series(rows, from_ts, bucket, _AGG_FIELDS)
+    return {"data": data, "from": from_ts, "to": to_ts, "meta": {"nics": nics}}

backend/tests/test_api_metrics.py

@@ -0,0 +1,74 @@
+"""Tests for /metrics read endpoints (Task 6)."""
+from datetime import datetime, timezone
+import time
+
+
+def test_metrics_endpoint_returns_series(client, admin_headers, db_session):
+    """Seeds InstanceMetric rows then reads them via the API.
+    The endpoint reads only the DB — no Proxmox call required."""
+    from app.models import InstanceMetric
+
+    now = int(time.time())
+    for i, cpu in enumerate((0.1, 0.3)):
+        db_session.add(InstanceMetric(
+            server_id=1, vmid=114, vm_type="qemu",
+            ts=datetime.fromtimestamp(now - 10 + i, tz=timezone.utc),
+            cpu=cpu, mem=100, maxmem=200, disk_used=10, disk_total=40,
+            netin_rate=1.0, netout_rate=2.0, diskread_rate=0.0, diskwrite_rate=0.0,
+            iops_read=5.0, iops_write=1.0))
+    db_session.commit()
+
+    r = client.get(
+        "/proxmox/api/1/vm/114/metrics?timeframe=hour&node=prod",
+        headers=admin_headers,
+    )
+    assert r.status_code == 200
+    body = r.json()
+    assert "data" in body and "meta" in body
+    assert body["data"], "expected at least one bucket"
+    assert "nics" in body["meta"]
+
+
+def test_container_metrics_endpoint_returns_series(client, admin_headers, db_session):
+    """Same flow for container endpoint."""
+    from app.models import InstanceMetric
+
+    now = int(time.time())
+    db_session.add(InstanceMetric(
+        server_id=1, vmid=200, vm_type="lxc",
+        ts=datetime.fromtimestamp(now - 5, tz=timezone.utc),
+        cpu=0.2, mem=512, maxmem=1024, disk_used=20, disk_total=80,
+        netin_rate=3.0, netout_rate=1.5, diskread_rate=0.0, diskwrite_rate=0.0,
+        iops_read=2.0, iops_write=0.5))
+    db_session.commit()
+
+    r = client.get(
+        "/proxmox/api/1/container/200/metrics?timeframe=hour&node=prod",
+        headers=admin_headers,
+    )
+    assert r.status_code == 200
+    body = r.json()
+    assert "data" in body and "meta" in body
+    assert "nics" in body["meta"]
+
+
+def test_metrics_explicit_from_to(client, admin_headers, db_session):
+    """Explicit from_ts/to_ts take precedence over timeframe."""
+    from app.models import InstanceMetric
+
+    now = int(time.time())
+    db_session.add(InstanceMetric(
+        server_id=1, vmid=300, vm_type="qemu",
+        ts=datetime.fromtimestamp(now - 30, tz=timezone.utc),
+        cpu=0.5, mem=256, maxmem=512, disk_used=5, disk_total=20,
+        netin_rate=0.5, netout_rate=0.5, diskread_rate=0.0, diskwrite_rate=0.0,
+        iops_read=1.0, iops_write=1.0))
+    db_session.commit()
+
+    r = client.get(
+        f"/proxmox/api/1/vm/300/metrics?from_ts={now - 60}&to_ts={now}",
+        headers=admin_headers,
+    )
+    assert r.status_code == 200
+    body = r.json()
+    assert "data" in body and "meta" in body

backend/tests/test_metrics_client.py

@@ -0,0 +1,37 @@
+from unittest.mock import MagicMock
+from app.proxmox.client import ProxmoxClient
+
+
+def _client_with_mock(proxmox):
+    c = ProxmoxClient.__new__(ProxmoxClient)
+    c.proxmox = proxmox
+    return c
+
+
+def test_get_vm_blockstats_returns_text():
+    px = MagicMock()
+    px.nodes.return_value.qemu.return_value.monitor.post.return_value = "drive-scsi0: rd_operations=5 wr_operations=2"
+    c = _client_with_mock(px)
+    assert "rd_operations=5" in c.get_vm_blockstats("prod", 114)
+    px.nodes.return_value.qemu.return_value.monitor.post.assert_called_once_with(command="info blockstats")
+
+
+def test_get_vm_blockstats_swallows_errors():
+    px = MagicMock()
+    px.nodes.return_value.qemu.return_value.monitor.post.side_effect = RuntimeError("nope")
+    c = _client_with_mock(px)
+    assert c.get_vm_blockstats("prod", 114) == ""
+
+
+def test_get_node_netstat_returns_rows():
+    px = MagicMock()
+    px.nodes.return_value.netstat.get.return_value = [{"dev": "tap114i0", "vmid": "114", "in": "1", "out": "2"}]
+    c = _client_with_mock(px)
+    assert c.get_node_netstat("prod")[0]["dev"] == "tap114i0"
+
+
+def test_get_node_netstat_swallows_errors():
+    px = MagicMock()
+    px.nodes.return_value.netstat.get.side_effect = RuntimeError("nope")
+    c = _client_with_mock(px)
+    assert c.get_node_netstat("prod") == []

backend/tests/test_metrics_collector.py

@@ -0,0 +1,83 @@
+"""Tests for metrics_collector.collect_once (unit-testable core)."""
+from unittest.mock import MagicMock
+from app.services.metrics_collector import collect_once
+
+
+def _client():
+    c = MagicMock()
+    # status/current
+    c.get_vm_status.return_value = {
+        "cpu": 0.5, "mem": 100, "maxmem": 200,
+        "netin": 1000, "netout": 2000, "diskread": 0, "diskwrite": 500,
+    }
+    c.get_vm_fsinfo.return_value = [{"used": 10, "total": 40}]
+    c.get_vm_blockstats.return_value = "drive-scsi0: rd_operations=100 wr_operations=50"
+    c.get_node_netstat.return_value = [
+        {"dev": "tap114i0", "vmid": "114", "in": "1000", "out": "2000"}]
+    return c
+
+
+def test_collect_once_first_tick_no_rates(db_session):
+    from app.models import InstanceMetric
+    items = [{"server_id": 1, "vmid": 114, "type": "qemu", "node": "prod", "status": "running"}]
+    clients = {1: _client()}
+    prev = {}
+    collect_once(db_session, items, clients, prev, _now=100.0)
+    db_session.commit()
+    m = db_session.query(InstanceMetric).one()
+    assert m.cpu == 0.5 and m.disk_used == 10 and m.disk_total == 40
+    assert m.netin_rate is None          # no previous counter yet
+    assert m.iops_read is None
+
+
+def test_collect_once_second_tick_computes_rates(db_session):
+    from app.models import InstanceMetric, InstanceNicMetric
+    items = [{"server_id": 1, "vmid": 114, "type": "qemu", "node": "prod", "status": "running"}]
+    c = _client()
+    clients = {1: c}
+    prev = {}
+    collect_once(db_session, items, clients, prev, _now=100.0)
+    # advance counters by 10 over 2 seconds
+    c.get_vm_status.return_value = {"cpu": 0.5, "mem": 100, "maxmem": 200,
+                                   "netin": 1020, "netout": 2040, "diskread": 0, "diskwrite": 500}
+    c.get_vm_blockstats.return_value = "drive-scsi0: rd_operations=110 wr_operations=56"
+    c.get_node_netstat.return_value = [{"dev": "tap114i0", "vmid": "114", "in": "1020", "out": "2040"}]
+    collect_once(db_session, items, clients, prev, _now=102.0)
+    db_session.commit()
+    rows = db_session.query(InstanceMetric).order_by(InstanceMetric.id).all()
+    assert rows[1].netin_rate == 10.0     # (1020-1000)/2
+    assert rows[1].iops_read == 5.0       # (110-100)/2
+    nic = db_session.query(InstanceNicMetric).order_by(InstanceNicMetric.id).all()[-1]
+    assert nic.in_rate == 10.0            # (1020-1000)/2
+
+
+def test_collect_once_evicts_stale_prev(db_session):
+    """Keys for VMs absent from the current tick must be pruned from prev."""
+    def _client_for(vmid_str):
+        c = MagicMock()
+        c.get_vm_status.return_value = {
+            "cpu": 0.1, "mem": 50, "maxmem": 100,
+            "netin": 0, "netout": 0, "diskread": 0, "diskwrite": 0,
+        }
+        c.get_vm_fsinfo.return_value = [{"used": 5, "total": 20}]
+        c.get_vm_blockstats.return_value = ""
+        c.get_node_netstat.return_value = [
+            {"dev": f"tap{vmid_str}i0", "vmid": vmid_str, "in": "0", "out": "0"}
+        ]
+        return c
+
+    prev = {}
+    # Tick 1: only vmid 114 running
+    items_114 = [{"server_id": 1, "vmid": 114, "type": "qemu", "node": "prod", "status": "running"}]
+    clients_114 = {1: _client_for("114")}
+    collect_once(db_session, items_114, clients_114, prev, _now=100.0)
+    # After tick 1, prev must contain 114 keys
+    assert any("114" in k for k in prev), "Expected 114 keys in prev after tick 1"
+
+    # Tick 2: only vmid 115 running (114 is gone)
+    items_115 = [{"server_id": 1, "vmid": 115, "type": "qemu", "node": "prod", "status": "running"}]
+    clients_115 = {1: _client_for("115")}
+    collect_once(db_session, items_115, clients_115, prev, _now=102.0)
+    # 114 keys must be evicted; 115 keys must be present
+    assert not any("114" in k for k in prev), "Expected 114 keys evicted from prev after tick 2"
+    assert any("115" in k for k in prev), "Expected 115 keys in prev after tick 2"

backend/tests/test_metrics_history.py

@@ -0,0 +1,71 @@
+from app.services import metrics_history as mh
+
+
+def test_compute_rate_normal():
+    assert mh.compute_rate(100, 160, 2.0) == 30.0
+
+def test_compute_rate_no_prev():
+    assert mh.compute_rate(None, 160, 2.0) is None
+
+def test_compute_rate_counter_reset():
+    assert mh.compute_rate(500, 10, 2.0) == 0.0
+
+def test_compute_rate_zero_dt():
+    assert mh.compute_rate(100, 200, 0) is None
+
+def test_parse_blockstats_sums_devices():
+    text = (
+        "drive-scsi0: rd_bytes=10 wr_bytes=20 rd_operations=100 wr_operations=50 flush_operations=1\n"
+        "drive-scsi1: rd_bytes=5 wr_bytes=5 rd_operations=7 wr_operations=3 flush_operations=0\n"
+    )
+    assert mh.parse_blockstats(text) == (107, 53)
+
+def test_parse_blockstats_empty():
+    assert mh.parse_blockstats("") == (0, 0)
+
+def test_parse_netstat_filters_by_vmid():
+    rows = [
+        {"dev": "tap114i0", "vmid": "114", "in": "1000", "out": "2000"},
+        {"dev": "tap114i1", "vmid": "114", "in": "10", "out": "20"},
+        {"dev": "tap200i0", "vmid": "200", "in": "9", "out": "9"},
+    ]
+    out = mh.parse_netstat(rows, 114)
+    assert out == {"tap114i0": (1000, 2000), "tap114i1": (10, 20)}
+
+def test_aggregate_fsinfo_sums():
+    disks = [{"used": 10, "total": 40}, {"used": 5, "total": 20}]
+    assert mh.aggregate_fsinfo(disks) == (15, 60)
+
+def test_aggregate_fsinfo_empty():
+    assert mh.aggregate_fsinfo([]) == (0, 0)
+
+
+def test_timeframe_to_range_hour():
+    assert mh.timeframe_to_range("hour", 10_000) == (10_000 - 3600, 10_000)
+
+def test_timeframe_to_range_month():
+    assert mh.timeframe_to_range("month", 10_000_000) == (10_000_000 - 30 * 86400, 10_000_000)
+
+def test_timeframe_to_range_unknown_defaults_hour():
+    assert mh.timeframe_to_range("nonsense", 10_000) == (10_000 - 3600, 10_000)
+
+def test_pick_bucket_respects_floor():
+    # 1h window, target 150 -> 24s ideal, but floor 15 keeps >=15
+    assert mh.pick_bucket_seconds(0, 3600, target_points=150, floor=15) == 24
+
+def test_pick_bucket_floor_applied():
+    assert mh.pick_bucket_seconds(0, 300, target_points=150, floor=15) == 15
+
+def test_aggregate_series_averages_buckets():
+    rows = [
+        {"time": 0, "cpu": 10.0, "mem": 100},
+        {"time": 5, "cpu": 20.0, "mem": 200},
+        {"time": 20, "cpu": 40.0, "mem": None},
+    ]
+    out = mh.aggregate_series(rows, from_ts=0, bucket=10, fields=["cpu", "mem"])
+    # bucket 0: times 0,5 -> cpu 15, mem 150 ; bucket 20: time 20 -> cpu 40, mem None
+    assert out[0] == {"time": 0, "cpu": 15.0, "mem": 150.0}
+    assert out[-1] == {"time": 20, "cpu": 40.0, "mem": None}
+
+def test_aggregate_series_empty():
+    assert mh.aggregate_series([], 0, 10, ["cpu"]) == []

backend/tests/test_metrics_models.py

@@ -0,0 +1,20 @@
+from datetime import datetime, timezone
+
+def test_instance_metric_roundtrip(db_session):
+    from app.models import InstanceMetric, InstanceNicMetric
+    ts = datetime(2026, 6, 21, 12, 0, tzinfo=timezone.utc)
+    db_session.add(InstanceMetric(
+        server_id=1, vmid=114, vm_type="qemu", ts=ts,
+        cpu=0.25, mem=100, maxmem=200, disk_used=10, disk_total=40,
+        netin_rate=1.0, netout_rate=2.0, diskread_rate=0.0, diskwrite_rate=3.0,
+        iops_read=12.0, iops_write=4.0,
+    ))
+    db_session.add(InstanceNicMetric(
+        server_id=1, vmid=114, ts=ts, dev="tap114i0", in_rate=5.0, out_rate=6.0))
+    db_session.commit()
+    m = db_session.query(InstanceMetric).one()
+    assert m.vmid == 114
+    assert m.iops_read == 12.0
+    n = db_session.query(InstanceNicMetric).one()
+    assert n.dev == "tap114i0"
+    assert n.out_rate == 6.0

backend/tests/test_metrics_query.py

@@ -0,0 +1,45 @@
+# backend/tests/test_metrics_query.py
+from datetime import datetime, timezone
+from app.services.metrics_history import query_instance_metrics
+
+
+def _row(db, InstanceMetric, ts, **kw):
+    db.add(InstanceMetric(server_id=1, vmid=114, vm_type="qemu",
+                          ts=datetime.fromtimestamp(ts, tz=timezone.utc), **kw))
+
+
+def test_query_aggregate_all_nic(db_session):
+    from app.models import InstanceMetric
+    _row(db_session, InstanceMetric, 1000, cpu=0.10, mem=100, maxmem=200,
+         disk_used=10, disk_total=40, netin_rate=1.0, netout_rate=2.0,
+         diskread_rate=0.0, diskwrite_rate=0.0, iops_read=5.0, iops_write=1.0)
+    _row(db_session, InstanceMetric, 1005, cpu=0.30, mem=140, maxmem=200,
+         disk_used=20, disk_total=40, netin_rate=3.0, netout_rate=4.0,
+         diskread_rate=0.0, diskwrite_rate=0.0, iops_read=7.0, iops_write=3.0)
+    db_session.commit()
+    out = query_instance_metrics(db_session, 1, 114, from_ts=1000, to_ts=1010, nic="all")
+    assert out["from"] == 1000 and out["to"] == 1010
+    pt = out["data"][0]
+    assert pt["cpu"] == 20.0          # avg(10%,30%)
+    assert pt["netin"] == 2.0         # avg(1,3)
+    assert pt["diskpct"] == 37.5      # avg(25%,50%)
+    assert pt["iops_read"] == 6.0
+
+
+def test_query_specific_nic_overrides_net(db_session):
+    from app.models import InstanceMetric, InstanceNicMetric
+    _row(db_session, InstanceMetric, 1000, cpu=0.0, mem=0, maxmem=1,
+         disk_used=0, disk_total=1, netin_rate=99.0, netout_rate=99.0,
+         diskread_rate=0.0, diskwrite_rate=0.0, iops_read=0.0, iops_write=0.0)
+    db_session.add(InstanceNicMetric(server_id=1, vmid=114,
+        ts=datetime.fromtimestamp(1000, tz=timezone.utc), dev="tap114i0",
+        in_rate=7.0, out_rate=8.0))
+    db_session.commit()
+    out = query_instance_metrics(db_session, 1, 114, 1000, 1010, nic="tap114i0")
+    assert out["meta"]["nics"] == ["tap114i0"]
+    assert out["data"][0]["netin"] == 7.0 and out["data"][0]["netout"] == 8.0
+
+
+def test_query_empty(db_session):
+    out = query_instance_metrics(db_session, 1, 999, 0, 10, nic="all")
+    assert out["data"] == [] and out["meta"]["nics"] == []

frontend/src/features/instances/tabs/GraphsTab.tsx

@@ -1,4 +1,4 @@
-import { useState } from 'react';
+import { useMemo, useState } from 'react';
 import { useTranslation } from 'react-i18next';
 import {
   Area,
@@ -12,7 +12,7 @@ import {
 import { Card, CardContent, CardHeader, CardTitle } from '@/components/ui/card';
 import { Button } from '@/components/ui/button';
 import { Loader2 } from 'lucide-react';
-import { useVMRrddata } from '@/hooks/use-instances';
+import { useVMMetrics } from '@/hooks/use-instances';
 import { formatBytes } from '@/lib/format';
 
 interface Props {
@@ -22,17 +22,31 @@ interface Props {
   node: string;
 }
 
-const TIMEFRAMES = [
-  { value: 'hour', label: '1h' },
-  { value: 'day', label: '24h' },
-  { value: 'week', label: '7d' },
-  { value: 'month', label: '30d' },
-  { value: 'year', label: '1y' },
+const PRESETS = [
+  { tf: 'hour', label: '1h' },
+  { tf: 'day', label: '24h' },
+  { tf: 'week', label: '7d' },
+  { tf: 'month', label: '30d' },
 ];
 
-function formatTime(ts: number): string {
+const MAX_SPAN = 30 * 86400;
+
+function fmtTime(ts: number): string {
   const d = new Date(ts * 1000);
-  return `${d.getHours().toString().padStart(2, '0')}:${d.getMinutes().toString().padStart(2, '0')}`;
+  return `${String(d.getHours()).padStart(2, '0')}:${String(d.getMinutes()).padStart(2, '0')}`;
+}
+
+// Convert a datetime-local string to unix seconds
+function localToUnix(val: string): number | undefined {
+  if (!val) return undefined;
+  return Math.floor(new Date(val).getTime() / 1000);
+}
+
+// Convert unix seconds to datetime-local value string
+function unixToLocal(ts: number): string {
+  const d = new Date(ts * 1000);
+  const pad = (n: number) => String(n).padStart(2, '0');
+  return `${d.getFullYear()}-${pad(d.getMonth() + 1)}-${pad(d.getDate())}T${pad(d.getHours())}:${pad(d.getMinutes())}`;
 }
 
 interface ChartCardProps {
@@ -42,13 +56,17 @@ interface ChartCardProps {
   color: string;
   formatValue?: (v: number) => string;
   unit?: string;
+  headerRight?: React.ReactNode;
 }
 
-function ChartCard({ title, data, dataKey, color, formatValue, unit }: ChartCardProps) {
+function ChartCard({ title, data, dataKey, color, formatValue, unit, headerRight }: ChartCardProps) {
   return (
     <Card>
       <CardHeader className="pb-2">
-        <CardTitle className="text-sm font-semibold">{title}</CardTitle>
+        <div className="flex items-center justify-between gap-2">
+          <CardTitle className="text-sm font-semibold">{title}</CardTitle>
+          {headerRight}
+        </div>
       </CardHeader>
       <CardContent>
         <ResponsiveContainer width="100%" height={200}>
@@ -62,7 +80,7 @@ function ChartCard({ title, data, dataKey, color, formatValue, unit }: ChartCard
             <CartesianGrid strokeDasharray="3 3" className="stroke-border" />
             <XAxis
               dataKey="time"
-              tickFormatter={formatTime}
+              tickFormatter={fmtTime}
               className="text-[10px] fill-muted-foreground"
               tick={{ fontSize: 10 }}
             />
@@ -70,17 +88,22 @@ function ChartCard({ title, data, dataKey, color, formatValue, unit }: ChartCard
               className="text-[10px] fill-muted-foreground"
               tick={{ fontSize: 10 }}
               tickFormatter={(v) => (formatValue ? formatValue(v) : String(v))}
-              width={50}
+              width={55}
             />
             <Tooltip
               contentStyle={{
-                backgroundColor: 'hsl(var(--popover))',
-                border: '1px solid hsl(var(--border))',
+                backgroundColor: 'rgba(0, 0, 0, 0.65)',
+                border: '1px solid rgba(255, 255, 255, 0.25)',
                 borderRadius: '6px',
                 fontSize: 12,
+                color: '#fff',
               }}
-              formatter={(val: any) => [
-                val != null && typeof val === 'number' && formatValue ? formatValue(val) : `${val ?? ''}${unit || ''}`,
+              labelStyle={{ color: '#fff' }}
+              itemStyle={{ color: '#fff' }}
+              formatter={(val: unknown) => [
+                val != null && typeof val === 'number' && formatValue
+                  ? formatValue(val)
+                  : `${val ?? ''}${unit || ''}`,
                 title,
               ]}
               labelFormatter={(ts) => new Date((ts as number) * 1000).toLocaleString()}
@@ -91,6 +114,7 @@ function ChartCard({ title, data, dataKey, color, formatValue, unit }: ChartCard
               stroke={color}
               strokeWidth={1.5}
               fill={`url(#grad-${dataKey})`}
+              dot={false}
             />
           </AreaChart>
         </ResponsiveContainer>
@@ -102,45 +126,115 @@ function ChartCard({ title, data, dataKey, color, formatValue, unit }: ChartCard
 export default function GraphsTab({ serverId, vmid, type, node }: Props) {
   const { t } = useTranslation();
   const [timeframe, setTimeframe] = useState('hour');
-  const { data: rrdData, isLoading } = useVMRrddata(serverId, vmid, type, node, timeframe);
+  // custom range: raw datetime-local strings
+  const [fromLocal, setFromLocal] = useState('');
+  const [toLocal, setToLocal] = useState('');
+  const [nic, setNic] = useState('all');
 
-  // Parse RRD data into chart-friendly format
-  const chartData = (rrdData as Array<Record<string, unknown>> | undefined)?.map((point) => ({
-    time: Number(point.time),
-    cpu: (Number(point.cpu) || 0) * 100,
-    mem: Number(point.mem) || Number(point.memused) || 0,
-    maxmem: Number(point.maxmem) || 0,
-    netin: Number(point.netin) || 0,
-    netout: Number(point.netout) || 0,
-    diskread: Number(point.diskread) || 0,
-    diskwrite: Number(point.diskwrite) || 0,
-  })) || [];
+  // Compute unix-second from/to with 30-day clamp
+  const customRange = useMemo<{ from?: number; to?: number }>(() => {
+    const f = localToUnix(fromLocal);
+    const to = localToUnix(toLocal);
+    if (f == null || to == null) return {};
+    let from = f;
+    if (to - from > MAX_SPAN) from = to - MAX_SPAN;
+    return { from, to };
+  }, [fromLocal, toLocal]);
+
+  const hasCustom = customRange.from != null && customRange.to != null;
+
+  const { data, isLoading } = useVMMetrics(serverId, vmid, type, node, {
+    timeframe: hasCustom ? undefined : timeframe,
+    from: customRange.from,
+    to: customRange.to,
+    nic,
+  });
+
+  const points = (data?.data ?? []) as Array<Record<string, unknown>>;
+  const nics = data?.meta?.nics ?? [];
+
+  function resetCustom() {
+    setFromLocal('');
+    setToLocal('');
+  }
+
+  // When the clamped from differs from the input, reflect it back
+  const displayFrom =
+    hasCustom && customRange.from != null ? unixToLocal(customRange.from) : fromLocal;
+
+  const nicSelector = (
+    <select
+      value={nic}
+      onChange={(e) => setNic(e.target.value)}
+      className="ml-1 rounded border border-border bg-background px-1 py-0.5 text-xs text-foreground focus:outline-none"
+    >
+      <option value="all">Все</option>
+      {nics.map((n) => (
+        <option key={n} value={n}>
+          {n}
+        </option>
+      ))}
+    </select>
+  );
 
   return (
     <div className="space-y-4">
-      {/* Timeframe selector */}
-      <div className="flex gap-1">
-        {TIMEFRAMES.map((tf) => (
-          <Button
-            key={tf.value}
-            variant={timeframe === tf.value ? 'default' : 'outline'}
-            size="sm"
-            onClick={() => setTimeframe(tf.value)}
-          >
-            {tf.label}
-          </Button>
-        ))}
+      {/* Time controls */}
+      <div className="flex flex-wrap items-center gap-2">
+        {/* Preset buttons */}
+        <div className="flex gap-1">
+          {PRESETS.map((p) => (
+            <Button
+              key={p.tf}
+              variant={!hasCustom && timeframe === p.tf ? 'default' : 'outline'}
+              size="sm"
+              onClick={() => {
+                setTimeframe(p.tf);
+                resetCustom();
+              }}
+            >
+              {p.label}
+            </Button>
+          ))}
+        </div>
+
+        {/* Custom range inputs */}
+        <div className="flex items-center gap-1 text-sm">
+          <input
+            type="datetime-local"
+            value={displayFrom}
+            onChange={(e) => setFromLocal(e.target.value)}
+            className="rounded border border-border bg-background px-2 py-1 text-xs text-foreground focus:outline-none"
+          />
+          <span className="text-muted-foreground">—</span>
+          <input
+            type="datetime-local"
+            value={toLocal}
+            onChange={(e) => setToLocal(e.target.value)}
+            className="rounded border border-border bg-background px-2 py-1 text-xs text-foreground focus:outline-none"
+          />
+          {hasCustom && (
+            <Button variant="outline" size="sm" onClick={resetCustom}>
+              Сбросить
+            </Button>
+          )}
+        </div>
       </div>
 
+      {/* Loading state */}
       {isLoading ? (
         <div className="flex items-center justify-center py-20">
           <Loader2 className="h-6 w-6 animate-spin text-muted-foreground" />
         </div>
+      ) : points.length === 0 ? (
+        <div className="flex items-center justify-center py-20 text-sm text-muted-foreground">
+          Нет данных
+        </div>
       ) : (
         <div className="grid gap-4 lg:grid-cols-2">
           <ChartCard
             title={t('graphs.cpu_usage')}
-            data={chartData}
+            data={points}
             dataKey="cpu"
             color="hsl(217, 91%, 60%)"
             formatValue={(v) => `${v.toFixed(1)}%`}
@@ -148,38 +242,64 @@ export default function GraphsTab({ serverId, vmid, type, node }: Props) {
           />
           <ChartCard
             title={t('graphs.memory_usage')}
-            data={chartData}
+            data={points}
             dataKey="mem"
             color="hsl(142, 76%, 36%)"
             formatValue={(v) => formatBytes(v)}
           />
           <ChartCard
-            title={t('graphs.network_in')}
-            data={chartData}
-            dataKey="netin"
-            color="hsl(262, 83%, 58%)"
-            formatValue={(v) => formatBytes(v)}
-          />
-          <ChartCard
-            title={t('graphs.network_out')}
-            data={chartData}
-            dataKey="netout"
-            color="hsl(25, 95%, 53%)"
-            formatValue={(v) => formatBytes(v)}
+            title={t('graphs.disk_write')}
+            data={points}
+            dataKey="diskwrite"
+            color="hsl(0, 84%, 60%)"
+            formatValue={(v) => `${formatBytes(v)}/s`}
           />
           <ChartCard
             title={t('graphs.disk_read')}
-            data={chartData}
+            data={points}
             dataKey="diskread"
             color="hsl(199, 89%, 48%)"
-            formatValue={(v) => formatBytes(v)}
+            formatValue={(v) => `${formatBytes(v)}/s`}
           />
           <ChartCard
-            title={t('graphs.disk_write')}
-            data={chartData}
-            dataKey="diskwrite"
-            color="hsl(0, 84%, 60%)"
-            formatValue={(v) => formatBytes(v)}
+            title={t('graphs.network_in')}
+            data={points}
+            dataKey="netin"
+            color="hsl(262, 83%, 58%)"
+            formatValue={(v) => `${formatBytes(v)}/s`}
+            headerRight={nicSelector}
+          />
+          <ChartCard
+            title={t('graphs.network_out')}
+            data={points}
+            dataKey="netout"
+            color="hsl(25, 95%, 53%)"
+            formatValue={(v) => `${formatBytes(v)}/s`}
+            headerRight={nicSelector}
+          />
+          <ChartCard
+            title="IOPS чтение"
+            data={points}
+            dataKey="iops_read"
+            color="hsl(180, 70%, 40%)"
+            formatValue={(v) => `${v.toFixed(0)} ops/s`}
+            unit=" ops/s"
+          />
+          <ChartCard
+            title="IOPS запись"
+            data={points}
+            dataKey="iops_write"
+            color="hsl(300, 60%, 50%)"
+            formatValue={(v) => `${v.toFixed(0)} ops/s`}
+            unit=" ops/s"
+          />
+          <ChartCard
+            title="Заполненность диска"
+            data={points}
+            dataKey="diskpct"
+            color="hsl(45, 93%, 47%)"
+            formatValue={(v) => `${v.toFixed(1)}%`}
+            unit="%"
           />
         </div>
       )}

frontend/src/features/instances/tabs/OverviewTab.tsx

@@ -1,7 +1,7 @@
 import { Cpu, MemoryStick, HardDrive, Clock, Wifi, Globe, Server, ArrowDown, ArrowUp } from 'lucide-react';
 import { Card, CardContent, CardHeader, CardTitle } from '@/components/ui/card';
 import { useVMLiveMetrics, useVMConfig, useVMInterfaces } from '@/hooks/use-instances';
-import { formatBytes, formatUptime, formatPercent } from '@/lib/format';
+import { formatBytes, formatUptime, formatPercent, vmVcpuCount } from '@/lib/format';
 import { useTranslation } from 'react-i18next';
 import type { DiskInfo } from '@/types';
 
@@ -165,7 +165,7 @@ export default function OverviewTab({ serverId, vmid, type, node }: Props) {
         <MetricCard
           label={t('common.cpu')}
           value={formatPercent(cpuPercent)}
-          subtitle={config?.cores ? `${config.cores} ${t('common.cores')}` : undefined}
+          subtitle={vmVcpuCount(config) ? `${vmVcpuCount(config)} ${t('common.cores')}` : undefined}
           percent={cpuPercent}
           icon={Cpu}
         />
@@ -258,7 +258,7 @@ export default function OverviewTab({ serverId, vmid, type, node }: Props) {
             <InfoRow label={t('common.vmid')} value={vmid} />
             <InfoRow label={t('common.node')} value={node} />
             <InfoRow label={t('common.type')} value={type === 'qemu' ? t('common.qemu_virtual_machine') : t('common.lxc_container')} />
-            <InfoRow label={t('common.vcpu')} value={config?.cores} />
+            <InfoRow label={t('common.vcpu')} value={vmVcpuCount(config)} />
             <InfoRow label={t('common.memory')} value={config?.memory ? `${config.memory} MB` : undefined} />
             <InfoRow label={t('common.os_type')} value={config?.ostype as string} />
             <InfoRow label={t('common.boot_order')} value={config?.boot as string} />

frontend/src/hooks/use-instances.ts

@@ -1,4 +1,4 @@
-import { useQuery, useMutation, useQueryClient } from '@tanstack/react-query';
+import { useQuery, useMutation, useQueryClient, keepPreviousData } from '@tanstack/react-query';
 import { useEffect, useState } from 'react';
 import { apiClient } from '@/lib/api-client';
 import { getTasksWebSocket } from '@/lib/websocket';
@@ -12,7 +12,6 @@ export const vmKeys = {
   detail: (serverId: number, vmid: number) => ['vm', serverId, vmid] as const,
   config: (serverId: number, vmid: number) => ['vm', serverId, vmid, 'config'] as const,
   status: (serverId: number, vmid: number) => ['vm', serverId, vmid, 'status'] as const,
-  rrddata: (serverId: number, vmid: number, tf: string) => ['vm', serverId, vmid, 'rrddata', tf] as const,
   snapshots: (serverId: number, vmid: number) => ['vm', serverId, vmid, 'snapshots'] as const,
   interfaces: (serverId: number, vmid: number) => ['vm', serverId, vmid, 'interfaces'] as const,
 };
@@ -178,14 +177,39 @@ export function useVMConfig(serverId: number, vmid: number, type: string, node:
   });
 }
 
-export function useVMRrddata(serverId: number, vmid: number, type: string, node: string, timeframe = 'hour') {
+
+export interface MetricPoint {
+  time: number;
+  cpu: number | null; mem: number | null; maxmem: number | null;
+  netin: number | null; netout: number | null;
+  diskread: number | null; diskwrite: number | null;
+  diskpct: number | null; iops_read: number | null; iops_write: number | null;
+}
+export interface MetricsResponse {
+  data: MetricPoint[]; from: number; to: number; meta: { nics: string[] };
+}
+
+export function useVMMetrics(
+  serverId: number, vmid: number, type: string, node: string,
+  opts: { timeframe?: string; from?: number; to?: number; nic?: string },
+) {
   const prefix = type === 'lxc' ? 'container' : 'vm';
-  return useQuery({
-    queryKey: vmKeys.rrddata(serverId, vmid, timeframe),
-    queryFn: async () => {
-      const res = await apiClient.get<{ data: unknown[]; timeframe: string }>(`/proxmox/api/${serverId}/${prefix}/${vmid}/rrddata?node=${node}&timeframe=${timeframe}`);
-      return res.data ?? [];
+  const { timeframe = 'hour', from, to, nic = 'all' } = opts;
+  return useQuery<MetricsResponse>({
+    queryKey: ['vm', serverId, vmid, 'metrics', { timeframe, from, to, nic }],
+    queryFn: () => {
+      const qs = new URLSearchParams({ node, nic });
+      if (from != null && to != null) {
+        qs.set('from_ts', String(from));
+        qs.set('to_ts', String(to));
+      } else {
+        qs.set('timeframe', timeframe);
+      }
+      return apiClient.get(`/proxmox/api/${serverId}/${prefix}/${vmid}/metrics?${qs}`);
     },
+    // Keep showing the previous series while a new range/NIC loads, so switching
+    // the interface selector doesn't blank the whole grid into a spinner.
+    placeholderData: keepPreviousData,
   });
 }
 
@@ -274,7 +298,11 @@ export function useSetVMOwner(serverId: number, vmid: number) {
   return useMutation({
     mutationFn: (user_id: number | null) =>
       apiClient.put(`/proxmox/api/${serverId}/vm/${vmid}/owner`, { user_id }),
-    onSuccess: () => qc.invalidateQueries({ queryKey: ['vm-owner', serverId, vmid] }),
+    onSuccess: () => {
+      qc.invalidateQueries({ queryKey: ['vm-owner', serverId, vmid] });
+      // Refresh the instances list so the owner column updates without a manual reload.
+      qc.invalidateQueries({ queryKey: vmKeys.all });
+    },
   });
 }
 

frontend/src/hooks/use-realtime-sync.ts

@@ -48,6 +48,24 @@ export function useGlobalRealtimeSync() {
       qc.invalidateQueries({ queryKey: vmKeys.resourcesAll });
     });
 
+    const unsubVmOwnerChanged = ws.onType('vm_owner_changed', (data: unknown) => {
+      const msg = data as {
+        vmid: number;
+        server_id: number;
+        owner_id: number | null;
+        owner_user: { username: string; email?: string; full_name?: string } | null;
+      };
+      qc.setQueryData<Array<Record<string, unknown>>>(vmKeys.all, (old) =>
+        old?.map((vm) =>
+          vm.server_id === msg.server_id && vm.vmid === msg.vmid
+            ? { ...vm, owner_id: msg.owner_id, owner_user: msg.owner_user }
+            : vm
+        )
+      );
+      // Keep the admin owner panel (settings tab) in sync as well.
+      qc.invalidateQueries({ queryKey: ['vm-owner', msg.server_id, msg.vmid] });
+    });
+
     // ── Server events ──────────────────────────────────────────────────────
     const invalidateServerCaches = () => {
       qc.invalidateQueries({ queryKey: nodeKeys.servers });
@@ -74,6 +92,7 @@ export function useGlobalRealtimeSync() {
       unsubVmStatus();
       unsubVmDeleted();
       unsubVmCreated();
+      unsubVmOwnerChanged();
       unsubServerAdded();
       unsubServerUpdated();
       unsubServerDeleted();

frontend/src/lib/format.ts

@@ -28,19 +28,32 @@ export function vmTypeLabel(type: string): string {
   return type === 'qemu' ? 'VM' : type === 'lxc' ? 'LXC' : type.toUpperCase();
 }
 
+/**
+ * Total vCPU count from a live Proxmox config.
+ * QEMU reports `cores` *per socket*, so the real total is `sockets * cores`
+ * (LXC has no sockets and reports the total in `cores`).
+ */
+export function vmVcpuCount(config?: { cores?: number; sockets?: number } | null): number | undefined {
+  if (!config?.cores) return undefined;
+  return config.cores * (config.sockets ?? 1);
+}
+
 /**
  * Formats a VM/LXC hardware configuration as "N vCPU / X GB RAM / Y GB".
- * `memory` is expected in MB; disk in GB (falls back to `maxdisk` in bytes).
+ * `memory` and `maxdisk` come from the cached instances list in bytes;
+ * `disk_size` (when present) is already in GB.
  */
 export function formatVmConfig(vm: {
   cores?: number;
   memory?: number;
+  maxmem?: number;
   disk_size?: number;
   maxdisk?: number;
 }): string {
   const parts: string[] = [];
   if (vm.cores) parts.push(`${vm.cores} vCPU`);
-  if (vm.memory) parts.push(`${Math.round(vm.memory / 1024)} GB RAM`);
+  const memBytes = vm.memory || vm.maxmem;
+  if (memBytes) parts.push(`${formatBytes(memBytes, 0)} RAM`);
   const diskGb = vm.disk_size ?? (vm.maxdisk ? Math.round(vm.maxdisk / 1024 ** 3) : 0);
   if (diskGb) parts.push(`${diskGb} GB`);
   return parts.join(' / ');

frontend/src/types/index.ts

@@ -120,6 +120,10 @@ export interface VMInstance {
 
 export interface VMConfig {
   cores?: number;
+  /** QEMU socket count; total vCPU = sockets * cores (defaults to 1) */
+  sockets?: number;
+  /** QEMU hotplugged vCPU count, when fewer than sockets * cores are online */
+  vcpus?: number;
   memory?: number;
   disk?: string;
   name?: string;