Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactor auto scaler logic #13

Merged
merged 4 commits into from
Aug 12, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
111 changes: 1 addition & 110 deletions cmd/internal/handle.go
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,6 @@ package internal
import (
"context"
"fmt"
"time"

"github.com/caarlos0/env/v9"
"golang.org/x/exp/slog"
Expand All @@ -17,117 +16,9 @@ func Handle(ctx context.Context, logger *slog.Logger) error {
return fmt.Errorf("could not parse environment variables: %w", err)
}

logger = logger.With(
"asg_arn", cfg.AutoscalingGroupARN,
"worker_pool_id", cfg.SpaceliftWorkerPoolID,
)

controller, err := internal.NewController(ctx, &cfg)
if err != nil {
return fmt.Errorf("could not create controller: %w", err)
}

workerPool, err := controller.GetWorkerPool(ctx)
if err != nil {
return fmt.Errorf("could not get worker pool: %w", err)
}

asg, err := controller.GetAutoscalingGroup(ctx)
if err != nil {
return fmt.Errorf("could not get autoscaling group: %w", err)
}

state, err := internal.NewState(workerPool, asg)
if err != nil {
return fmt.Errorf("could not create state: %w", err)
}

// Let's make sure that for each of the in-service instances we have a
// corresponding worker in Spacelift, or that we have "stray" machines.
if strayInstances := state.StrayInstances(); len(strayInstances) > 0 {
// There's a question of what to do with the "stray" machines. The
// decision will be made based on the creation timestamp.
instances, err := controller.DescribeInstances(ctx, strayInstances)
if err != nil {
return fmt.Errorf("could not list EC2 instances: %w", err)
}

for _, instance := range instances {
logger = logger.With("instance_id", *instance.InstanceId)
instanceAge := time.Since(*instance.LaunchTime)

logger = logger.With(
"launch_timestamp", instance.LaunchTime.Unix(),
"instance_age", instanceAge,
)

// If the machine was only created recently (say a generous window of 10
// minutes), it is possible that it hasn't managed to register itself with
// Spacelift yet. But if it's been around for a while we will want to kill
// it and remove it from the ASG.
if instanceAge > 10*time.Minute {
logger.Warn("instance has no corresponding worker in Spacelift, removing from the ASG")

if err := controller.KillInstance(ctx, *instance.InstanceId); err != nil {
return fmt.Errorf("could not kill instance: %w", err)
}

// We don't want to kill too many instances at once, so let's
// return after the first successfully killed one.
logger.Info("instance successfully removed from the ASG and terminated")

return nil
}
}
}

decision := state.Decide(cfg.AutoscalingMaxCreate, cfg.AutoscalingMaxKill)

if decision.ScalingDirection == internal.ScalingDirectionNone {
logger.Info("no scaling decision to be made")
return nil
}

if decision.ScalingDirection == internal.ScalingDirectionUp {
logger.With("instances", decision.ScalingSize).Info("scaling up the ASG")

if err := controller.ScaleUpASG(ctx, *asg.DesiredCapacity+int32(decision.ScalingSize)); err != nil {
return fmt.Errorf("could not scale up ASG: %w", err)
}

return nil
}

// If we got this far, we're scaling down.
logger.With("instances", decision.ScalingSize).Info("scaling down ASG")

idleWorkers := state.IdleWorkers()

for i := 0; i < decision.ScalingSize; i++ {
worker := idleWorkers[i]

_, instanceID, _ := worker.InstanceIdentity()

logger = logger.With(
"worker_id", worker.ID,
"instance_id", instanceID,
)
logger.Info("scaling down ASG and killing worker")

drained, err := controller.DrainWorker(ctx, worker.ID)
if err != nil {
return fmt.Errorf("could not drain worker: %w", err)
}

if !drained {
logger.Warn("worker was busy, stopping the scaling down process")
return nil
}

if err := controller.KillInstance(ctx, string(instanceID)); err != nil {
return fmt.Errorf("could not kill instance: %w", err)
}
}

return nil
return internal.NewAutoScaler(controller, logger).Scale(ctx, cfg)
}
141 changes: 141 additions & 0 deletions internal/auto_scaler.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,141 @@
package internal

import (
"context"
"fmt"
"time"

autoscalingtypes "github.com/aws/aws-sdk-go-v2/service/autoscaling/types"
ec2types "github.com/aws/aws-sdk-go-v2/service/ec2/types"
"golang.org/x/exp/slog"
)

//go:generate mockery --output ./ --name ControllerInterface --filename mock_controller_test.go --outpkg internal_test --structname MockController
type ControllerInterface interface {
DescribeInstances(ctx context.Context, instanceIDs []string) (instances []ec2types.Instance, err error)
GetAutoscalingGroup(ctx context.Context) (out *autoscalingtypes.AutoScalingGroup, err error)
GetWorkerPool(ctx context.Context) (out *WorkerPool, err error)
DrainWorker(ctx context.Context, workerID string) (drained bool, err error)
KillInstance(ctx context.Context, instanceID string) (err error)
ScaleUpASG(ctx context.Context, desiredCapacity int32) (err error)
}

type AutoScaler struct {
controller ControllerInterface
logger *slog.Logger
}

func NewAutoScaler(controller ControllerInterface, logger *slog.Logger) *AutoScaler {
return &AutoScaler{controller: controller, logger: logger}
}

func (s AutoScaler) Scale(ctx context.Context, cfg RuntimeConfig) error {
logger := s.logger.With(
"asg_arn", cfg.AutoscalingGroupARN,
"worker_pool_id", cfg.SpaceliftWorkerPoolID,
)

workerPool, err := s.controller.GetWorkerPool(ctx)
if err != nil {
return fmt.Errorf("could not get worker pool: %w", err)
}

asg, err := s.controller.GetAutoscalingGroup(ctx)
if err != nil {
return fmt.Errorf("could not get autoscaling group: %w", err)
}

state, err := NewState(workerPool, asg)
if err != nil {
return fmt.Errorf("could not create state: %w", err)
}

// Let's make sure that for each of the in-service instances we have a
// corresponding worker in Spacelift, or that we have "stray" machines.
if strayInstances := state.StrayInstances(); len(strayInstances) > 0 {
// There's a question of what to do with the "stray" machines. The
// decision will be made based on the creation timestamp.
instances, err := s.controller.DescribeInstances(ctx, strayInstances)
if err != nil {
return fmt.Errorf("could not list EC2 instances: %w", err)
}

for _, instance := range instances {
logger = logger.With("instance_id", *instance.InstanceId)
instanceAge := time.Since(*instance.LaunchTime)

logger = logger.With(
"launch_timestamp", instance.LaunchTime.Unix(),
"instance_age", instanceAge,
)

// If the machine was only created recently (say a generous window of 10
// minutes), it is possible that it hasn't managed to register itself with
// Spacelift yet. But if it's been around for a while we will want to kill
// it and remove it from the ASG.
if instanceAge > 10*time.Minute {
logger.Warn("instance has no corresponding worker in Spacelift, removing from the ASG")

if err := s.controller.KillInstance(ctx, *instance.InstanceId); err != nil {
return fmt.Errorf("could not kill instance: %w", err)
}

// We don't want to kill too many instances at once, so let's
// return after the first successfully killed one.
logger.Info("instance successfully removed from the ASG and terminated")

return nil
}
}
}

decision := state.Decide(cfg.AutoscalingMaxCreate, cfg.AutoscalingMaxKill)

if decision.ScalingDirection == ScalingDirectionNone {
logger.Info("no scaling decision to be made")
return nil
}

if decision.ScalingDirection == ScalingDirectionUp {
logger.With("instances", decision.ScalingSize).Info("scaling up the ASG")

if err := s.controller.ScaleUpASG(ctx, *asg.DesiredCapacity+int32(decision.ScalingSize)); err != nil {
return fmt.Errorf("could not scale up ASG: %w", err)
}

return nil
}

// If we got this far, we're scaling down.
logger.With("instances", decision.ScalingSize).Info("scaling down ASG")

idleWorkers := state.IdleWorkers()

for i := 0; i < decision.ScalingSize; i++ {
worker := idleWorkers[i]

_, instanceID, _ := worker.InstanceIdentity()

logger = logger.With(
"worker_id", worker.ID,
"instance_id", instanceID,
)
logger.Info("scaling down ASG and killing worker")

drained, err := s.controller.DrainWorker(ctx, worker.ID)
if err != nil {
return fmt.Errorf("could not drain worker: %w", err)
}

if !drained {
logger.Warn("worker was busy, stopping the scaling down process")
return nil
}

if err := s.controller.KillInstance(ctx, string(instanceID)); err != nil {
return fmt.Errorf("could not kill instance: %w", err)
}
}

return nil
}
122 changes: 122 additions & 0 deletions internal/auto_scaler_test.go
Original file line number Diff line number Diff line change
@@ -0,0 +1,122 @@
package internal_test

import (
"bytes"
"context"
"testing"

"github.com/aws/aws-sdk-go-v2/service/autoscaling/types"
"github.com/stretchr/testify/mock"
"github.com/stretchr/testify/require"
"golang.org/x/exp/slog"

"github.com/spacelift-io/awsautoscalr/internal"
)

func TestAutoScalerScalingNone(t *testing.T) {
var buf bytes.Buffer
h := slog.NewTextHandler(&buf, nil)

cfg := internal.RuntimeConfig{}

ctrl := new(MockController)
defer ctrl.AssertExpectations(t)

scaler := internal.NewAutoScaler(ctrl, slog.New(h))

ctrl.On("GetWorkerPool", mock.Anything).Return(&internal.WorkerPool{
Workers: []internal.Worker{
{
ID: "1",
Metadata: `{"asg_id": "group", "instance_id": "instance"}`,
},
},
}, nil)
ctrl.On("GetAutoscalingGroup", mock.Anything).Return(&types.AutoScalingGroup{
AutoScalingGroupName: ptr("group"),
MinSize: ptr(int32(1)),
MaxSize: ptr(int32(3)),
DesiredCapacity: ptr(int32(2)),
}, nil)
err := scaler.Scale(context.Background(), cfg)
require.NoError(t, err)
}

func TestAutoScalerScalingUp(t *testing.T) {
var buf bytes.Buffer
h := slog.NewTextHandler(&buf, nil)

cfg := internal.RuntimeConfig{}

ctrl := new(MockController)
defer ctrl.AssertExpectations(t)

scaler := internal.NewAutoScaler(ctrl, slog.New(h))

ctrl.On("GetWorkerPool", mock.Anything).Return(&internal.WorkerPool{
Workers: []internal.Worker{
{
ID: "1",
Metadata: `{"asg_id": "group", "instance_id": "instance"}`,
},
},
PendingRuns: 2,
}, nil)
ctrl.On("GetAutoscalingGroup", mock.Anything).Return(&types.AutoScalingGroup{
AutoScalingGroupName: ptr("group"),
MinSize: ptr(int32(1)),
MaxSize: ptr(int32(3)),
DesiredCapacity: ptr(int32(2)),
Instances: []types.Instance{
{},
},
}, nil)
ctrl.On("ScaleUpASG", mock.Anything, int32(2)).Return(nil)
err := scaler.Scale(context.Background(), cfg)
require.NoError(t, err)
}

func TestAutoScalerScalingDown(t *testing.T) {
var buf bytes.Buffer
h := slog.NewTextHandler(&buf, nil)

cfg := internal.RuntimeConfig{
AutoscalingMaxKill: 1,
}

ctrl := new(MockController)
defer ctrl.AssertExpectations(t)

scaler := internal.NewAutoScaler(ctrl, slog.New(h))

ctrl.On("GetWorkerPool", mock.Anything).Return(&internal.WorkerPool{
Workers: []internal.Worker{
{
ID: "1",
Metadata: `{"asg_id": "group", "instance_id": "instance"}`,
},
{
ID: "2",
Metadata: `{"asg_id": "group", "instance_id": "instance2"}`,
},
},
}, nil)
ctrl.On("GetAutoscalingGroup", mock.Anything).Return(&types.AutoScalingGroup{
AutoScalingGroupName: ptr("group"),
MinSize: ptr(int32(1)),
MaxSize: ptr(int32(3)),
DesiredCapacity: ptr(int32(2)),
Instances: []types.Instance{
{},
{},
},
}, nil)
ctrl.On("DrainWorker", mock.Anything, "1").Return(true, nil)
ctrl.On("KillInstance", mock.Anything, "instance").Return(nil)
err := scaler.Scale(context.Background(), cfg)
require.NoError(t, err)
}

func ptr[T any](v T) *T {
return &v
}
Loading
Loading