Overhaul wal for create workload

cluster/calcium/create.go

@@ -55,14 +55,33 @@ func (c *Calcium) doCreateWorkloads(ctx context.Context, opts *types.DeployOptio
 		defer func() {
 			cctx, cancel := context.WithTimeout(utils.InheritTracingInfo(ctx, context.TODO()), c.config.GlobalTimeout)
 			for nodename := range deployMap {
-				if e := c.store.DeleteProcessing(cctx, opts.GetProcessing(nodename)); e != nil {
+				processing := opts.GetProcessing(nodename)
+				if e := c.store.DeleteProcessing(cctx, processing); e != nil {
 					logger.Errorf(ctx, "[Calcium.doCreateWorkloads] delete processing failed for %s: %+v", nodename, e)
 				}
 			}
 			close(ch)
 			cancel()
 		}()
 
+		var resourceCommit wal.Commit
+		defer func() {
+			if resourceCommit != nil {
+				if err := resourceCommit(); err != nil {
+					logger.Errorf(ctx, "commit wal failed: %s, %+v", eventWorkloadResourceAllocated, err)
+				}
+			}
+		}()
+
+		var processingCommits map[string]wal.Commit
+		defer func() {
+			for nodename := range processingCommits {
+				if err := processingCommits[nodename](); err != nil {
+					logger.Errorf(ctx, "commit wal failed: %s, %s, %+v", eventProcessingCreated, nodename, err)
+				}
+			}
+		}()
+
 		_ = utils.Txn(
 			ctx,
 
@@ -81,15 +100,23 @@ func (c *Calcium) doCreateWorkloads(ctx context.Context, opts *types.DeployOptio
 
 					// commit changes
 					nodes := []*types.Node{}
+					processingCommits = make(map[string]wal.Commit)
 					for nodename, deploy := range deployMap {
 						for _, plan := range plans {
 							plan.ApplyChangesOnNode(nodeMap[nodename], utils.Range(deploy)...)
 						}
 						nodes = append(nodes, nodeMap[nodename])
-						if err = c.store.CreateProcessing(ctx, opts.GetProcessing(nodename), deploy); err != nil {
+						processing := opts.GetProcessing(nodename)
+						if processingCommits[nodename], err = c.wal.Log(eventProcessingCreated, processing); err != nil {
+							return errors.WithStack(err)
+						}
+						if err = c.store.CreateProcessing(ctx, processing, deploy); err != nil {
 							return errors.WithStack(err)
 						}
 					}
+					if resourceCommit, err = c.wal.Log(eventWorkloadResourceAllocated, nodes); err != nil {
+						return errors.WithStack(err)
+					}
 					return errors.WithStack(c.store.UpdateNodes(ctx, nodes...))
 				})
 			},
@@ -246,6 +273,7 @@ func (c *Calcium) doDeployOneWorkload(
 	config *enginetypes.VirtualizationCreateOptions,
 	decrProcessing bool,
 ) (err error) {
+	logger := log.WithField("Calcium", "doDeployWorkload").WithField("nodename", node.Name).WithField("opts", opts).WithField("msg", msg)
 	workload := &types.Workload{
 		ResourceMeta: types.ResourceMeta{
 			CPU:               msg.CPU,
@@ -276,7 +304,7 @@ func (c *Calcium) doDeployOneWorkload(
 	defer func() {
 		if commit != nil {
 			if err := commit(); err != nil {
-				log.Errorf(ctx, "[doDeployOneWorkload] Commit WAL %s failed: %v", eventCreateWorkload, err)
+				logger.Errorf(ctx, "Commit WAL %s failed: %+v", eventWorkloadCreated, err)
 			}
 		}
 	}()
@@ -292,10 +320,11 @@ func (c *Calcium) doDeployOneWorkload(
 			// We couldn't WAL the workload ID above VirtualizationCreate temporarily,
 			// so there's a time gap window, once the core process crashes between
 			// VirtualizationCreate and logCreateWorkload then the worload is leaky.
-			if commit, err = c.wal.logCreateWorkload(workload.ID, node.Name); err != nil {
-				return err
-			}
-			return nil
+			commit, err = c.wal.Log(eventWorkloadCreated, &types.Workload{
+				ID:       workload.ID,
+				Nodename: workload.Nodename,
+			})
+			return errors.WithStack(err)
 		},
 
 		func(ctx context.Context) (err error) {
@@ -375,13 +404,13 @@ func (c *Calcium) doDeployOneWorkload(
 
 		// remove workload
 		func(ctx context.Context, _ bool) error {
-			log.Errorf(ctx, "[doDeployOneWorkload] failed to deploy workload %s, rollback", workload.ID)
+			logger.Errorf(ctx, "[doDeployOneWorkload] failed to deploy workload %s, rollback", workload.ID)
 			if workload.ID == "" {
 				return nil
 			}
 
 			if err := c.store.RemoveWorkload(ctx, workload); err != nil {
-				log.Errorf(ctx, "[doDeployOneWorkload] failed to remove workload %s")
+				logger.Errorf(ctx, "[doDeployOneWorkload] failed to remove workload %s", workload.ID)
 			}
 
 			return workload.Remove(ctx, true)

cluster/calcium/create_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"testing"
 
+	"github.com/pkg/errors"
 	enginemocks "github.com/projecteru2/core/engine/mocks"
 	enginetypes "github.com/projecteru2/core/engine/types"
 	lockmocks "github.com/projecteru2/core/lock/mocks"
@@ -15,8 +16,6 @@ import (
 	"github.com/projecteru2/core/types"
 	"github.com/projecteru2/core/wal"
 	walmocks "github.com/projecteru2/core/wal/mocks"
-
-	"github.com/pkg/errors"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/mock"
 )
@@ -107,8 +106,12 @@ func TestCreateWorkloadTxn(t *testing.T) {
 
 	c.wal = &WAL{WAL: &walmocks.WAL{}}
 	mwal := c.wal.WAL.(*walmocks.WAL)
-	defer mwal.AssertExpectations(t)
 	var walCommitted bool
+	commit := wal.Commit(func() error {
+		walCommitted = true
+		return nil
+	})
+	mwal.On("Log", mock.Anything, mock.Anything).Return(commit, nil)
 
 	store.On("CreateProcessing", mock.Anything, mock.Anything, mock.Anything).Return(nil)
 	store.On("DeleteProcessing", mock.Anything, mock.Anything, mock.Anything).Return(nil)
@@ -191,7 +194,7 @@ func TestCreateWorkloadTxn(t *testing.T) {
 	assert.EqualValues(t, 1, node2.CPUUsed)
 	node1.CPUUsed = 0
 	node2.CPUUsed = 0
-	assert.False(t, walCommitted)
+	assert.True(t, walCommitted)
 
 	// doCreateWorkloadOnNode fails: doGetAndPrepareNode
 	store.On("UpdateNodes", mock.Anything, mock.Anything, mock.Anything).Return(nil)
@@ -224,13 +227,13 @@ func TestCreateWorkloadTxn(t *testing.T) {
 	assert.EqualValues(t, 2, cnt)
 	assert.EqualValues(t, 0, node1.CPUUsed)
 	assert.EqualValues(t, 0, node2.CPUUsed)
-	assert.False(t, walCommitted)
+	assert.True(t, walCommitted)
 
 	// doDeployOneWorkload fails: VirtualizationCreate
 	engine.On("ImageLocalDigests", mock.Anything, mock.Anything).Return([]string{""}, nil)
 	engine.On("ImageRemoteDigest", mock.Anything, mock.Anything).Return("", nil)
 	engine.On("VirtualizationCreate", mock.Anything, mock.Anything).Return(nil, errors.Wrap(context.DeadlineExceeded, "VirtualizationCreate")).Twice()
-	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(nil)
+	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(1, nil)
 	store.On("ListNodeWorkloads", mock.Anything, mock.Anything, mock.Anything).Return(nil, types.ErrNoETCD)
 	walCommitted = false
 	ch, err = c.CreateWorkload(ctx, opts)
@@ -245,18 +248,13 @@ func TestCreateWorkloadTxn(t *testing.T) {
 	assert.EqualValues(t, 2, cnt)
 	assert.EqualValues(t, 0, node1.CPUUsed)
 	assert.EqualValues(t, 0, node2.CPUUsed)
-	assert.False(t, walCommitted)
+	assert.True(t, walCommitted)
 
 	// doCreateAndStartWorkload fails: AddWorkload
 	engine.On("VirtualizationCreate", mock.Anything, mock.Anything).Return(&enginetypes.VirtualizationCreated{ID: "c1"}, nil)
 	engine.On("VirtualizationStart", mock.Anything, mock.Anything).Return(nil)
 	engine.On("VirtualizationInspect", mock.Anything, mock.Anything).Return(&enginetypes.VirtualizationInfo{}, nil)
 	store.On("AddWorkload", mock.Anything, mock.Anything, mock.Anything).Return(errors.Wrap(context.DeadlineExceeded, "AddWorkload")).Twice()
-	commit := wal.Commit(func() error {
-		walCommitted = true
-		return nil
-	})
-	mwal.On("Log", eventCreateWorkload, mock.Anything).Return(commit, nil)
 	walCommitted = false
 	ch, err = c.CreateWorkload(ctx, opts)
 	assert.Nil(t, err)

cluster/calcium/lambda_test.go

@@ -29,7 +29,8 @@ func TestRunAndWaitFailedThenWALCommitted(t *testing.T) {
 	c.wal = &WAL{WAL: &walmocks.WAL{}}
 
 	mwal := c.wal.WAL.(*walmocks.WAL)
-	defer mwal.AssertExpectations(t)
+	defer mwal.AssertNotCalled(t, "Log")
+	mwal.On("Log", mock.Anything, mock.Anything).Return(nil, nil)
 
 	opts := &types.DeployOptions{
 		Name:           "zc:name",
@@ -267,7 +268,7 @@ func newLambdaCluster(t *testing.T) (*Calcium, []*types.Node) {
 	// doCreateAndStartWorkload fails: AddWorkload
 	engine.On("VirtualizationCreate", mock.Anything, mock.Anything).Return(&enginetypes.VirtualizationCreated{ID: "workloadfortonictest"}, nil)
 	engine.On("VirtualizationStart", mock.Anything, mock.Anything).Return(nil)
-	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(nil)
+	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(1, nil)
 	store.On("ListNodeWorkloads", mock.Anything, mock.Anything, mock.Anything).Return(nil, types.ErrNoETCD)
 	engine.On("VirtualizationInspect", mock.Anything, mock.Anything).Return(&enginetypes.VirtualizationInfo{}, nil)
 	store.On("AddWorkload", mock.Anything, mock.Anything, mock.Anything).Return(nil)

cluster/calcium/remove_test.go

@@ -65,7 +65,7 @@ func TestRemoveWorkload(t *testing.T) {
 	// success
 	engine := &enginemocks.API{}
 	workload.Engine = engine
-	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(nil)
+	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(1, nil)
 	store.On("GetWorkloads", mock.Anything, mock.Anything).Return([]*types.Workload{workload}, nil)
 	store.On("RemoveWorkload", mock.Anything, mock.Anything).Return(nil)
 	store.On("UpdateNodeResource", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(nil)

cluster/calcium/replace_test.go

@@ -169,7 +169,7 @@ func TestReplaceWorkload(t *testing.T) {
 	engine.On("VirtualizationCreate", mock.Anything, mock.Anything).Return(nil, types.ErrCannotGetEngine).Once()
 	engine.On("VirtualizationStart", mock.Anything, mock.Anything).Return(types.ErrCannotGetEngine).Once()
 	//store.On("UpdateNodeResource", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(nil).Once()
-	//engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(nil).Once()
+	//engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(1, nil).Once()
 	//store.On("RemoveWorkload", mock.Anything, mock.Anything).Return(nil).Once()
 	ch, err = c.ReplaceWorkload(ctx, opts)
 	assert.NoError(t, err)
@@ -200,7 +200,7 @@ func TestReplaceWorkload(t *testing.T) {
 	store.AssertExpectations(t)
 	engine.AssertExpectations(t)
 
-	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(nil)
+	engine.On("VirtualizationRemove", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(1, nil)
 	store.On("RemoveWorkload", mock.Anything, mock.Anything).Return(nil)
 	store.On("ListNodeWorkloads", mock.Anything, mock.Anything, mock.Anything).Return(nil, types.ErrNoETCD)
 	// succ