feat: initial watchdog implementation (#1341)

* feat: initial watchdog implementation

Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com>

* fiuxups

* Add more output

* wip: idletime checker

* wire idle watchdog checks

* enlarge watchdog time window

* small fixes

* Use stopmodel

* Always delete process

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

---------

Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com>
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

api/api.go

@@ -13,6 +13,7 @@ import (
 	"github.com/go-skynet/LocalAI/internal"
 	"github.com/go-skynet/LocalAI/metrics"
 	"github.com/go-skynet/LocalAI/pkg/assets"
+	"github.com/go-skynet/LocalAI/pkg/model"
 
 	"github.com/gofiber/fiber/v2"
 	"github.com/gofiber/fiber/v2/middleware/cors"
@@ -79,6 +80,22 @@ func Startup(opts ...options.AppOption) (*options.Option, *config.ConfigLoader,
 		options.Loader.StopAllGRPC()
 	}()
 
+	if options.WatchDog {
+		wd := model.NewWatchDog(
+			options.Loader,
+			options.WatchDogBusyTimeout,
+			options.WatchDogIdleTimeout,
+			options.WatchDogBusy,
+			options.WatchDogIdle)
+		options.Loader.SetWatchDog(wd)
+		go wd.Run()
+		go func() {
+			<-options.Context.Done()
+			log.Debug().Msgf("Context canceled, shutting down")
+			wd.Shutdown()
+		}()
+	}
+
 	return options, cl, nil
 }
 

api/localai/backend_monitor.go

@@ -128,7 +128,7 @@ func BackendMonitorEndpoint(bm BackendMonitor) func(c *fiber.Ctx) error {
 			return fmt.Errorf("backend %s is not currently loaded", backendId)
 		}
 
-		status, rpcErr := model.GRPC(false).Status(context.TODO())
+		status, rpcErr := model.GRPC(false, nil).Status(context.TODO())
 		if rpcErr != nil {
 			log.Warn().Msgf("backend %s experienced an error retrieving status info: %s", backendId, rpcErr.Error())
 			val, slbErr := bm.SampleLocalBackendProcess(backendId)

api/options/options.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"embed"
 	"encoding/json"
+	"time"
 
 	"github.com/go-skynet/LocalAI/metrics"
 	"github.com/go-skynet/LocalAI/pkg/gallery"
@@ -38,6 +39,11 @@ type Option struct {
 
 	SingleBackend           bool
 	ParallelBackendRequests bool
+
+	WatchDogIdle                             bool
+	WatchDogBusy                             bool
+	WatchDog                                 bool
+	WatchDogBusyTimeout, WatchDogIdleTimeout time.Duration
 }
 
 type AppOption func(*Option)
@@ -63,6 +69,32 @@ func WithCors(b bool) AppOption {
 	}
 }
 
+var EnableWatchDog = func(o *Option) {
+	o.WatchDog = true
+}
+
+var EnableWatchDogIdleCheck = func(o *Option) {
+	o.WatchDog = true
+	o.WatchDogIdle = true
+}
+
+var EnableWatchDogBusyCheck = func(o *Option) {
+	o.WatchDog = true
+	o.WatchDogBusy = true
+}
+
+func SetWatchDogBusyTimeout(t time.Duration) AppOption {
+	return func(o *Option) {
+		o.WatchDogBusyTimeout = t
+	}
+}
+
+func SetWatchDogIdleTimeout(t time.Duration) AppOption {
+	return func(o *Option) {
+		o.WatchDogIdleTimeout = t
+	}
+}
+
 var EnableSingleBackend = func(o *Option) {
 	o.SingleBackend = true
 }