Dynamic PDB to Game Room Namespaces

cmd/runtimewatcher/wire.go

@@ -31,6 +31,7 @@ import (
 	"github.com/topfreegames/maestro/internal/core/services/events"
 	"github.com/topfreegames/maestro/internal/core/services/workers"
 	"github.com/topfreegames/maestro/internal/core/worker"
+	workerconfigs "github.com/topfreegames/maestro/internal/core/worker/config"
 	"github.com/topfreegames/maestro/internal/core/worker/runtimewatcher"
 	"github.com/topfreegames/maestro/internal/service"
 )
@@ -42,16 +43,24 @@ func provideRuntimeWatcherBuilder() *worker.WorkerBuilder {
 	}
 }
 
+func provideRuntimeWatcherConfig(c config.Config) *workerconfigs.RuntimeWatcherConfig {
+	return &workerconfigs.RuntimeWatcherConfig{
+		DisruptionWorkerIntervalSeconds: c.GetDuration("runtimeWatcher.disruptionWorker.intervalSeconds"),
+		DisruptionSafetyPercentage:      c.GetFloat64("runtimeWatcher.disruptionWorker.safetyPercentage"),
+	}
+}
+
 var WorkerOptionsSet = wire.NewSet(
 	service.NewRuntimeKubernetes,
+	service.NewRoomStorageRedis,
 	RoomManagerSet,
-	wire.Struct(new(worker.WorkerOptions), "RoomManager", "Runtime"))
+	provideRuntimeWatcherConfig,
+	wire.Struct(new(worker.WorkerOptions), "Runtime", "RoomStorage", "RoomManager", "RuntimeWatcherConfig"))
 
 var RoomManagerSet = wire.NewSet(
 	service.NewSchedulerStoragePg,
 	service.NewClockTime,
 	service.NewPortAllocatorRandom,
-	service.NewRoomStorageRedis,
 	service.NewGameRoomInstanceStorageRedis,
 	service.NewSchedulerCacheRedis,
 	service.NewRoomManagerConfig,

config/config.yaml

@@ -59,6 +59,10 @@ workers:
 reporter:
   metrics:
     intervalMillis: 10000
+runtimeWatcher:
+  disruptionWorker:
+    intervalSeconds: 5
+    safetyPercentage: 0.05
 
 services:
   roomManager:

docs/reference/Architecture.md

@@ -46,7 +46,8 @@ You could find all operations at [Operations section](Operations.md#available-op
 
 > Note: In Maestro a worker is a collection of routines that executes a flow related to one and only one **Scheduler** each.
 
-Runtime Watcher Worker listens to runtime events related to the **Scheduler** and reflects the changes in **Maestro**. Currently, it listens for Game Rooms creation, deletion, and update.
+Runtime Watcher Worker listens to runtime events related to the **Scheduler** and reflects the changes in **Maestro**. Currently, it mitigate disruptions by looking at the current
+amount of occupied rooms, and it listens for Game Rooms creation, deletion, and update.
 
 ![Runtime Watcher Worker IMAGE](../images/Architecture-Runtime-Watcher-Worker.jpg)
 

docs/reference/Kubernetes.md

@@ -39,8 +39,31 @@ flowchart BT
 ```
 
 ### Runtime watcher
-The runtime watcher component maintains a worker process for each scheduler that keeps watching and processing _change
-events_ in pods resources. For doing that, it uses a [pods informer](https://pkg.go.dev/k8s.io/client-go/informers),
+The runtime watcher component spawn two types of workers: one that is responsible for mitigating disruptions and another for processing _change events_ in pods resources.
+
+#### Disruption Worker
+
+This worker consists in a single goroutine with a ticker. Each time it runs, it will
+check the number of occupied rooms at the time and try to mitigate disruptions. For k8s,
+this mitigation consists on applying a PDB to the scheduler's namespace, that has
+`minAvailable` equals to the number of occupied rooms plus a safety percentage.
+
+One can configure the interval in which this worker runs and also the safety percentage
+in `config/config.yaml`:
+
+```yaml
+runtimeWatcher:
+  disruptionWorker:
+    intervalSeconds: 5
+    safetyPercentage: 0.05
+```
+
+#### Pod Change Events Worker
+
+For this type of worker, runtime watcher spawns multiple goroutines, maintaining a worker
+process for each scheduler that keeps watching and processing _change events_ in pods
+resources. For doing that, it uses a
+[pods informer](https://pkg.go.dev/k8s.io/client-go/informers),
 binding handlers for **add**, **update** and **delete** events for all pods managed by it.
 
 This component is not responsible for updating/creating/deleting

internal/adapters/runtime/kubernetes/kubernetes.go

@@ -23,16 +23,22 @@
 package kubernetes
 
 import (
+	"github.com/topfreegames/maestro/internal/core/logs"
 	"github.com/topfreegames/maestro/internal/core/ports"
+	"go.uber.org/zap"
 	kube "k8s.io/client-go/kubernetes"
 )
 
 var _ ports.Runtime = (*kubernetes)(nil)
 
 type kubernetes struct {
 	clientSet kube.Interface
+	logger    *zap.Logger
 }
 
 func New(clientSet kube.Interface) *kubernetes {
-	return &kubernetes{clientSet}
+	return &kubernetes{
+		clientSet: clientSet,
+		logger:    zap.L().With(zap.String(logs.LogFieldRuntime, "kubernetes")),
+	}
 }

internal/adapters/runtime/kubernetes/scheduler.go

@@ -24,14 +24,113 @@ package kubernetes
 
 import (
 	"context"
+	"strconv"
 
 	"github.com/topfreegames/maestro/internal/core/entities"
 	"github.com/topfreegames/maestro/internal/core/ports/errors"
+	"go.uber.org/zap"
 	v1 "k8s.io/api/core/v1"
+	v1Policy "k8s.io/api/policy/v1"
 	kerrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/intstr"
 )
 
+const (
+	DefaultDisruptionSafetyPercentage float64 = 0.05
+	MajorKubeVersionPDB               int     = 1
+	MinorKubeVersionPDB               int     = 21
+)
+
+func (k *kubernetes) isPDBSupported() bool {
+	// Check based on the kube version of the clientSet if PDBs are supported (1.21+)
+	version, err := k.clientSet.Discovery().ServerVersion()
+	if err != nil {
+		k.logger.Warn("Could not get kube API version, can not check for PDB support", zap.Error(err))
+		return false
+	}
+	major, err := strconv.Atoi(version.Major)
+	if err != nil {
+		k.logger.Warn(
+			"Could not convert major kube API version to int, can not check for PDB support",
+			zap.String("majorKubeAPIVersion", version.Major),
+		)
+		return false
+	}
+	if major < MajorKubeVersionPDB {
+		k.logger.Warn(
+			"Can not create PDB for this kube API version",
+			zap.Int("majorKubeAPIVersion", major),
+			zap.Int("majorPDBVersionRequired", MajorKubeVersionPDB),
+		)
+		return false
+	}
+	minor, err := strconv.Atoi(version.Minor)
+	if err != nil {
+		k.logger.Warn(
+			"Could not convert minor kube API version to int, can not check for PDB support",
+			zap.String("minorKubeAPIVersion", version.Minor),
+		)
+		return false
+	}
+	if minor < MinorKubeVersionPDB {
+		k.logger.Warn(
+			"Can not create PDB for this kube API version",
+			zap.Int("minorKubeAPIVersion", minor),
+			zap.Int("minorPDBVersionRequired", MinorKubeVersionPDB),
+		)
+		return false
+	}
+	return true
+}
+
+func (k *kubernetes) createPDBFromScheduler(ctx context.Context, scheduler *entities.Scheduler) (*v1Policy.PodDisruptionBudget, error) {
+	if scheduler == nil {
+		return nil, errors.NewErrInvalidArgument("scheduler pointer can not be nil")
+	}
+	pdbSpec := &v1Policy.PodDisruptionBudget{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: scheduler.Name,
+		},
+		Spec: v1Policy.PodDisruptionBudgetSpec{
+			MinAvailable: &intstr.IntOrString{
+				Type:   intstr.Int,
+				IntVal: int32(0),
+			},
+		},
+	}
+
+	if scheduler.Autoscaling != nil && scheduler.Autoscaling.Enabled {
+		pdbSpec.Spec.MinAvailable = &intstr.IntOrString{
+			Type:   intstr.Int,
+			IntVal: int32(scheduler.Autoscaling.Min),
+		}
+	}
+
+	pdb, err := k.clientSet.PolicyV1().PodDisruptionBudgets(scheduler.Name).Create(ctx, pdbSpec, metav1.CreateOptions{})
+	if err != nil && !kerrors.IsAlreadyExists(err) {
+		k.logger.Warn("error creating pdb", zap.String("scheduler", scheduler.Name), zap.Error(err))
+		return nil, err
+	}
+
+	return pdb, nil
+}
+
+func (k *kubernetes) deletePDBFromScheduler(ctx context.Context, scheduler *entities.Scheduler) error {
+	if scheduler == nil {
+		return errors.NewErrInvalidArgument("scheduler pointer can not be nil")
+	}
+	if !k.isPDBSupported() {
+		return errors.NewErrUnexpected("PDBs are not supported for this kube API version")
+	}
+	err := k.clientSet.PolicyV1().PodDisruptionBudgets(scheduler.Name).Delete(ctx, scheduler.Name, metav1.DeleteOptions{})
+	if err != nil && !kerrors.IsNotFound(err) {
+		k.logger.Warn("error deleting pdb", zap.String("scheduler", scheduler.Name), zap.Error(err))
+		return err
+	}
+	return nil
+}
+
 func (k *kubernetes) CreateScheduler(ctx context.Context, scheduler *entities.Scheduler) error {
 	namespace := &v1.Namespace{
 		ObjectMeta: metav1.ObjectMeta{
@@ -48,11 +147,20 @@ func (k *kubernetes) CreateScheduler(ctx context.Context, scheduler *entities.Sc
 		return errors.NewErrUnexpected("error creating scheduler: %s", err)
 	}
 
+	_, err = k.createPDBFromScheduler(ctx, scheduler)
+	if err != nil {
+		k.logger.Warn("PDB Creation during scheduler creation failed", zap.String("scheduler", scheduler.Name), zap.Error(err))
+	}
+
 	return nil
 }
 
 func (k *kubernetes) DeleteScheduler(ctx context.Context, scheduler *entities.Scheduler) error {
-	err := k.clientSet.CoreV1().Namespaces().Delete(ctx, scheduler.Name, metav1.DeleteOptions{})
+	err := k.deletePDBFromScheduler(ctx, scheduler)
+	if err != nil {
+		k.logger.Warn("PDB Deletion during scheduler deletion failed", zap.String("scheduler", scheduler.Name), zap.Error(err))
+	}
+	err = k.clientSet.CoreV1().Namespaces().Delete(ctx, scheduler.Name, metav1.DeleteOptions{})
 	if err != nil {
 		if kerrors.IsNotFound(err) {
 			return errors.NewErrNotFound("scheduler '%s' not found", scheduler.Name)
@@ -63,3 +171,57 @@ func (k *kubernetes) DeleteScheduler(ctx context.Context, scheduler *entities.Sc
 
 	return nil
 }
+
+func (k *kubernetes) MitigateDisruption(
+	ctx context.Context,
+	scheduler *entities.Scheduler,
+	roomAmount int,
+	safetyPercentage float64,
+) error {
+	if scheduler == nil {
+		return errors.NewErrInvalidArgument("empty pointer received for scheduler, can not mitigate disruptions")
+	}
+
+	incSafetyPercentage := 1.0
+	if safetyPercentage < DefaultDisruptionSafetyPercentage {
+		k.logger.Warn(
+			"invalid safety percentage, using default percentage",
+			zap.Float64("safetyPercentage", safetyPercentage),
+			zap.Float64("DefaultDisruptionSafetyPercentage", DefaultDisruptionSafetyPercentage),
+		)
+		safetyPercentage = DefaultDisruptionSafetyPercentage
+	}
+	incSafetyPercentage += safetyPercentage
+
+	// For kubernetes mitigating disruptions means updating the current PDB
+	// minAvailable to the number of occupied rooms if above a threshold
+	pdb, err := k.clientSet.PolicyV1().PodDisruptionBudgets(scheduler.Name).Get(ctx, scheduler.Name, metav1.GetOptions{})
+	if err != nil && !kerrors.IsNotFound(err) {
+		// Non-recoverable errors
+		return errors.NewErrUnexpected("non recoverable error when getting PDB for scheduler '%s': %s", scheduler.Name, err)
+	}
+
+	if pdb == nil || kerrors.IsNotFound(err) {
+		pdb, err = k.createPDBFromScheduler(ctx, scheduler)
+		if err != nil {
+			return errors.NewErrUnexpected("error creating PDB for scheduler '%s': %s", scheduler.Name, err)
+		}
+	}
+
+	currentPdbMinAvailable := pdb.Spec.MinAvailable.IntVal
+	if currentPdbMinAvailable == int32(float64(roomAmount)*incSafetyPercentage) {
+		return nil
+	}
+
+	pdb.Spec.MinAvailable = &intstr.IntOrString{
+		Type:   intstr.Int,
+		IntVal: int32(float64(roomAmount) * incSafetyPercentage),
+	}
+
+	_, err = k.clientSet.PolicyV1().PodDisruptionBudgets(scheduler.Name).Update(ctx, pdb, metav1.UpdateOptions{})
+	if err != nil {
+		return errors.NewErrUnexpected("error updating PDB to mitigate disruptions for scheduler '%s': %s", scheduler.Name, err)
+	}
+
+	return nil
+}