felhom-agent/internal/hub/collect.go

package hub

import (
	"context"
	"fmt"
	"log/slog"
	"time"

	"gitea.dooplex.hu/admin/felhom-agent/internal/proxmox"
)

// proxmoxReader is the read-only subset the collector needs. Signatures match the
// REAL internal/proxmox.Client surface (slice 1): the node is held by the Client
// (no per-call node arg), reads return values (not pointers), and the guest type is
// proxmox.Guest. (The task's sketch used node-arg/pointer/LXC shapes; adapted to
// the actual exports per its instruction — no proxmox changes were needed: ListLXC
// already carries status/maxmem/maxdisk, GuestConfig carries cores.)
type proxmoxReader interface {
	Node() string
	NodeStatus(ctx context.Context) (proxmox.NodeStatus, error)
	ListLXC(ctx context.Context) ([]proxmox.Guest, error)
	GuestConfig(ctx context.Context, vmid int) (proxmox.GuestConfig, error)
}

// Collector builds a HostReport from read-only sources. All deps are behind narrow
// interfaces for unit testing.
type Collector struct {
	px           proxmoxReader
	cf           CloudflaredProber
	hostID       string
	agentVersion string
	logger       *slog.Logger
	now          func() time.Time
}

// NewCollector builds a collector. hostID echoes config.Hub.HostID; agentVersion is
// the binary version.
func NewCollector(px proxmoxReader, cf CloudflaredProber, hostID, agentVersion string, logger *slog.Logger) *Collector {
	if logger == nil {
		logger = slog.Default()
	}
	return &Collector{
		px:           px,
		cf:           cf,
		hostID:       hostID,
		agentVersion: agentVersion,
		logger:       logger,
		now:          func() time.Time { return time.Now().UTC() },
	}
}

// Collect builds the report. Best-effort liveness: a failed NodeStatus is a hard
// error (no useful report — the cycle skips the POST); a failed per-guest
// GuestConfig degrades that guest to status="unknown" without spec but still sends;
// a cloudflared probe failure yields status="unknown" and is never fatal.
func (c *Collector) Collect(ctx context.Context) (*HostReport, error) {
	ns, err := c.px.NodeStatus(ctx)
	if err != nil {
		return nil, fmt.Errorf("hub: NodeStatus failed (no useful report): %w", err)
	}

	report := &HostReport{
		HostID:       c.hostID,
		ReportedAt:   c.now().Format(time.RFC3339),
		AgentVersion: c.agentVersion,
		Host:         hostMetrics(c.px.Node(), ns),
		Guests:       c.collectGuests(ctx),
		// Defined-but-empty this slice (slices 5/6). Non-nil so they marshal as [].
		StorageTargets: []StorageTarget{},
		Backups:        []Backup{},
		RestoreTests:   []RestoreTest{},
		PBSSnapshots:   []PBSSnapshot{},
		AuditTail:      []AuditEntry{},
		Cloudflared:    Cloudflared{Status: c.cloudflaredStatus(ctx)},
	}
	return report, nil
}

func hostMetrics(node string, ns proxmox.NodeStatus) HostMetrics {
	h := HostMetrics{
		Node:             node,
		CPUPercent:       ns.CPU * 100, // PVE cpu is a 0..1 fraction
		MemoryTotalBytes: ns.Memory.Total,
		MemoryUsedBytes:  ns.Memory.Used,
		DiskTotalBytes:   ns.RootFS.Total,
		DiskUsedBytes:    ns.RootFS.Used,
		LoadAvg:          ns.LoadAvg,
		UptimeSeconds:    ns.Uptime,
	}
	h.MemoryPercent = percent(ns.Memory.Used, ns.Memory.Total)
	h.DiskPercent = percent(ns.RootFS.Used, ns.RootFS.Total)
	if h.LoadAvg == nil {
		h.LoadAvg = []string{}
	}
	return h
}

func (c *Collector) collectGuests(ctx context.Context) []Guest {
	lxc, err := c.px.ListLXC(ctx)
	if err != nil {
		// Not fatal: a report with no guest list still carries host liveness.
		c.logger.Warn("hub: ListLXC failed; reporting no guests", "err", err)
		return []Guest{}
	}
	guests := make([]Guest, 0, len(lxc))
	for _, g := range lxc {
		entry := Guest{VMID: g.VMID, Name: g.Name, Status: g.Status, ControllerVersion: ""}
		// Normalize an empty run-status to "unknown" so the wire value is always one
		// of running|stopped|unknown (matches the hub handler's empty→unknown default).
		if entry.Status == "" {
			entry.Status = "unknown"
		}
		// GuestConfig supplies cores; memory/disk come from the list entry (bytes).
		// On failure, KEEP the known run-status from ListLXC — only the spec is lost.
		cfg, err := c.px.GuestConfig(ctx, g.VMID)
		if err != nil {
			c.logger.Warn("hub: GuestConfig failed; spec omitted (run-status kept)",
				"vmid", g.VMID, "err", err)
			entry.Spec = nil
		} else {
			entry.Spec = &GuestSpec{
				Cores:       cfg.Cores,
				MemoryBytes: g.MaxMem,
				DiskBytes:   g.MaxDisk,
			}
		}
		guests = append(guests, entry)
	}
	return guests
}

func (c *Collector) cloudflaredStatus(ctx context.Context) string {
	if c.cf == nil {
		return "unknown"
	}
	st, err := c.cf.Status(ctx)
	if err != nil || st == "" {
		c.logger.Warn("hub: cloudflared probe failed", "err", err)
		return "unknown"
	}
	return st
}

func percent(used, total int64) float64 {
	if total <= 0 {
		return 0
	}
	return float64(used) / float64(total) * 100
}