ananke/cmd/hecate/main.go

package main

import (
	"context"
	"errors"
	"flag"
	"fmt"
	"log"
	"os"
	"os/signal"
	"syscall"
	"time"

	"scm.bstein.dev/bstein/hecate/internal/cluster"
	"scm.bstein.dev/bstein/hecate/internal/config"
	"scm.bstein.dev/bstein/hecate/internal/execx"
	"scm.bstein.dev/bstein/hecate/internal/service"
	"scm.bstein.dev/bstein/hecate/internal/state"
	"scm.bstein.dev/bstein/hecate/internal/ups"
)

func main() {
	logger := log.New(os.Stdout, "[hecate] ", log.LstdFlags)
	if len(os.Args) < 2 {
		usage()
		os.Exit(2)
	}

	sub := os.Args[1]
	switch sub {
	case "startup":
		if err := runStartup(logger, os.Args[2:]); err != nil {
			logger.Printf("startup failed: %v", err)
			os.Exit(1)
		}
	case "shutdown":
		if err := runShutdown(logger, os.Args[2:]); err != nil {
			logger.Printf("shutdown failed: %v", err)
			os.Exit(1)
		}
	case "daemon":
		if err := runDaemon(logger, os.Args[2:]); err != nil {
			logger.Printf("daemon failed: %v", err)
			os.Exit(1)
		}
	case "status":
		if err := runStatus(logger, os.Args[2:]); err != nil {
			logger.Printf("status failed: %v", err)
			os.Exit(1)
		}
	case "help", "-h", "--help":
		usage()
	default:
		logger.Printf("unknown command: %s", sub)
		usage()
		os.Exit(2)
	}
}

func runStartup(logger *log.Logger, args []string) error {
	fs := flag.NewFlagSet("startup", flag.ExitOnError)
	configPath := fs.String("config", "/etc/hecate/hecate.yaml", "Path to config file")
	execute := fs.Bool("execute", false, "Actually execute changes (default dry-run)")
	forceBranch := fs.String("force-flux-branch", "", "Patch Flux source branch before resume")
	skipLocalBootstrap := fs.Bool("skip-local-bootstrap", false, "Skip local fallback bootstrap applies")
	_ = fs.Parse(args)

	_, orch, err := buildOrchestrator(logger, *configPath, !*execute)
	if err != nil {
		return err
	}
	ctx, cancel := context.WithCancel(context.Background())
	defer cancel()

	return orch.Startup(ctx, cluster.StartupOptions{
		ForceFluxBranch:    *forceBranch,
		SkipLocalBootstrap: *skipLocalBootstrap,
		Reason:             "manual-startup",
	})
}

func runShutdown(logger *log.Logger, args []string) error {
	fs := flag.NewFlagSet("shutdown", flag.ExitOnError)
	configPath := fs.String("config", "/etc/hecate/hecate.yaml", "Path to config file")
	execute := fs.Bool("execute", false, "Actually execute changes (default dry-run)")
	skipEtcd := fs.Bool("skip-etcd-snapshot", false, "Skip etcd snapshot")
	skipDrain := fs.Bool("skip-drain", false, "Skip worker drain")
	reason := fs.String("reason", "manual-shutdown", "Shutdown reason for run history")
	_ = fs.Parse(args)

	_, orch, err := buildOrchestrator(logger, *configPath, !*execute)
	if err != nil {
		return err
	}
	ctx, cancel := context.WithCancel(context.Background())
	defer cancel()

	return orch.Shutdown(ctx, cluster.ShutdownOptions{
		SkipEtcdSnapshot: *skipEtcd,
		SkipDrain:        *skipDrain,
		Reason:           *reason,
	})
}

func runDaemon(logger *log.Logger, args []string) error {
	fs := flag.NewFlagSet("daemon", flag.ExitOnError)
	configPath := fs.String("config", "/etc/hecate/hecate.yaml", "Path to config file")
	dryRunActions := fs.Bool("dry-run-actions", false, "Log planned actions without executing")
	_ = fs.Parse(args)

	cfg, orch, err := buildOrchestrator(logger, *configPath, *dryRunActions)
	if err != nil {
		return err
	}
	if !cfg.UPS.Enabled {
		return fmt.Errorf("UPS monitoring is disabled in config")
	}
	targets := make([]service.Target, 0, len(cfg.UPS.Targets)+1)
	switch cfg.UPS.Provider {
	case "nut":
		if len(cfg.UPS.Targets) == 0 {
			target := cfg.UPS.Target
			if target == "" {
				return fmt.Errorf("ups target must be set")
			}
			targets = append(targets, service.Target{
				Name:     "primary",
				Target:   target,
				Provider: ups.NewNUTProvider(target),
			})
		} else {
			for idx, t := range cfg.UPS.Targets {
				name := t.Name
				if name == "" {
					name = fmt.Sprintf("target-%d", idx+1)
				}
				targets = append(targets, service.Target{
					Name:     name,
					Target:   t.Target,
					Provider: ups.NewNUTProvider(t.Target),
				})
			}
		}
	default:
		return fmt.Errorf("unsupported UPS provider: %s", cfg.UPS.Provider)
	}

	d := service.NewDaemon(cfg, orch, targets, logger)
	ctx, stop := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
	defer stop()
	if err := d.Run(ctx); err != nil && !errors.Is(err, context.Canceled) {
		return err
	}
	return nil
}

func runStatus(logger *log.Logger, args []string) error {
	fs := flag.NewFlagSet("status", flag.ExitOnError)
	configPath := fs.String("config", "/etc/hecate/hecate.yaml", "Path to config file")
	_ = fs.Parse(args)

	cfg, orch, err := buildOrchestrator(logger, *configPath, true)
	if err != nil {
		return err
	}
	recs, err := state.New(cfg.State.RunHistoryPath).Load()
	if err != nil {
		return err
	}
	last := "none"
	if len(recs) > 0 {
		r := recs[len(recs)-1]
		last = fmt.Sprintf("%s success=%t duration=%ds at=%s", r.Action, r.Success, r.DurationSeconds, r.EndedAt.Format(time.RFC3339))
	}
	logger.Printf("expected_flux_branch=%s", cfg.ExpectedFluxBranch)
	logger.Printf("control_planes=%v", cfg.ControlPlanes)
	logger.Printf("estimated_shutdown_budget_seconds=%d", orch.EstimatedShutdownSeconds())
	logger.Printf("last_run=%s", last)
	return nil
}

func buildOrchestrator(logger *log.Logger, cfgPath string, dryRun bool) (config.Config, *cluster.Orchestrator, error) {
	cfg, err := config.Load(cfgPath)
	if err != nil {
		return config.Config{}, nil, err
	}
	if err := state.EnsureDir(cfg.State.Dir); err != nil {
		return config.Config{}, nil, err
	}
	runner := &execx.Runner{
		DryRun:     dryRun,
		Kubeconfig: cfg.Kubeconfig,
		Logger:     logger,
	}
	store := state.New(cfg.State.RunHistoryPath)
	orch := cluster.New(cfg, runner, store, logger)
	return cfg, orch, nil
}

func usage() {
	fmt.Print(`hecate: staged startup/shutdown + UPS-triggered protection

Usage:
  hecate <command> [flags]

Commands:
  startup   Perform staged cluster startup
  shutdown  Perform graceful cluster shutdown
  daemon    Monitor UPS and auto-trigger shutdown
  status    Print current hecate status and estimates

Examples:
  hecate startup --config /etc/hecate/hecate.yaml --execute --force-flux-branch main
  hecate shutdown --config /etc/hecate/hecate.yaml --execute --reason "manual-maintenance"
  hecate daemon --config /etc/hecate/hecate.yaml
  hecate status --config /etc/hecate/hecate.yaml
`)
}