chore: refactor error types and add units

add a detailed error

fix tests

cluster/calcium/cluster.go

@@ -1,7 +1,6 @@
 package calcium
 
 import (
-	"fmt"
 	"strings"
 
 	"github.com/projecteru2/core/cluster"
@@ -52,7 +51,7 @@ func New(config types.Config) (*Calcium, error) {
 	case cluster.GITHUB:
 		scm = github.New(config)
 	default:
-		return nil, fmt.Errorf("Unkonwn SCM type: %s", config.Git.SCMType)
+		return nil, types.NewDetailedErr(types.ErrBadSCMType, scmtype)
 	}
 
 	return &Calcium{store: store, config: config, scheduler: scheduler, network: titanium, source: scm}, nil

cluster/calcium/common.go

@@ -8,7 +8,7 @@ import (
 
 const (
 	restartAlways = "always"
-	minMemory     = 4194304
+	minMemory     = types.MByte * 4
 	root          = "root"
 	maxPuller     = 10
 )

cluster/calcium/create_container.go

@@ -31,23 +31,24 @@ func (c *Calcium) CreateContainer(ctx context.Context, opts *types.DeployOptions
 
 	// 4194304 Byte = 4 MB, docker 创建容器的内存最低标准
 	if opts.Memory < minMemory {
-		return nil, fmt.Errorf("Minimum memory limit allowed is 4MB, got %d", opts.Memory)
+		return nil, types.NewDetailedErr(types.ErrBadMemory,
+			fmt.Sprintf("Minimum memory limit allowed is 4MB, got %d", opts.Memory))
 	}
 	// Count 要大于0
 	if opts.Count <= 0 {
-		return nil, fmt.Errorf("Count must be positive, got %d", opts.Count)
+		return nil, types.NewDetailedErr(types.ErrBadCount, opts.Count)
 	}
 	// CPUQuota 也需要大于 0
 	if opts.CPUQuota <= 0 {
-		return nil, fmt.Errorf("CPUQuota must be positive, got %d", opts.Count)
+		return nil, types.NewDetailedErr(types.ErrBadCPU, opts.CPUQuota)
 	}
 
 	if pod.Favor == scheduler.MEMORY_PRIOR {
 		return c.createContainerWithMemoryPrior(ctx, opts)
 	} else if pod.Favor == scheduler.CPU_PRIOR {
 		return c.createContainerWithCPUPrior(ctx, opts)
 	}
-	return nil, fmt.Errorf("[CreateContainer] Favor not support: %v", pod.Favor)
+	return nil, types.NewDetailedErr(types.ErrBadFaver, pod.Favor)
 }
 
 func (c *Calcium) createContainerWithMemoryPrior(ctx context.Context, opts *types.DeployOptions) (chan *types.CreateContainerMessage, error) {
@@ -121,7 +122,7 @@ func (c *Calcium) createContainerWithCPUPrior(ctx context.Context, opts *types.D
 	}
 
 	if len(nodesInfo) == 0 {
-		return ch, fmt.Errorf("Not enough resource to create container")
+		return ch, types.ErrInsufficientRes
 	}
 
 	go func() {
@@ -304,7 +305,7 @@ func (c *Calcium) makeContainerOptions(index int, quota types.CPUMap, opts *type
 	} else if favor == scheduler.MEMORY_PRIOR {
 		resource = makeMemoryPriorSetting(opts.Memory, opts.CPUQuota, opts.SoftLimit)
 	} else {
-		return nil, nil, nil, "", fmt.Errorf("favor not support %s", favor)
+		return nil, nil, nil, "", types.NewDetailedErr(types.ErrBadFaver, favor)
 	}
 	resource.Ulimits = ulimits
 

cluster/calcium/create_container_test.go

@@ -68,8 +68,7 @@ func TestClean(t *testing.T) {
 
 	// delete pod, which will fail because there are remaining nodes
 	err := mockc.store.RemovePod(context.Background(), podname)
-	assert.Error(t, err)
-	assert.Contains(t, err.Error(), "still has nodes")
+	assert.EqualError(t, types.IsDetailedErr(err), types.ErrPodHasNodes.Error())
 }
 
 func TestCreateContainerWithCPUPrior(t *testing.T) {

cluster/calcium/network.go

@@ -2,7 +2,6 @@ package calcium
 
 import (
 	"context"
-	"fmt"
 
 	"github.com/projecteru2/core/types"
 	"github.com/projecteru2/core/utils"
@@ -20,7 +19,7 @@ func (c *Calcium) ListNetworks(ctx context.Context, podname string, driver strin
 	}
 
 	if len(nodes) == 0 {
-		return networks, fmt.Errorf("Pod %s has no nodes", podname)
+		return networks, types.NewDetailedErr(types.ErrPodNoNodes, podname)
 	}
 
 	node := nodes[0]

cluster/calcium/realloc.go

@@ -2,7 +2,6 @@ package calcium
 
 import (
 	"context"
-	"fmt"
 	"sync"
 
 	enginecontainer "github.com/docker/docker/api/types/container"
@@ -57,10 +56,10 @@ func (c *Calcium) ReallocResource(ctx context.Context, IDs []string, cpu float64
 			newCPURequire := container.Quota + cpu
 			newMemRequire := container.Memory + mem
 			if newCPURequire < 0 {
-				return nil, fmt.Errorf("CPU can not below zero %v", newCPURequire)
+				return nil, types.NewDetailedErr(types.ErrBadCPU, newCPURequire)
 			}
 			if newMemRequire < minMemory {
-				return nil, fmt.Errorf("Memory can not below zero %v", newMemRequire)
+				return nil, types.NewDetailedErr(types.ErrBadMemory, newMemRequire)
 			}
 
 			// init data
@@ -115,7 +114,7 @@ func (c *Calcium) reallocNodesMemory(ctx context.Context, podname string, nodeCo
 	defer lock.Unlock(ctx)
 	for node, containers := range nodeContainers {
 		if cap := int(node.MemCap / memory); cap < len(containers) {
-			return fmt.Errorf("Not enough resource %s", node.Name)
+			return types.NewDetailedErr(types.ErrInsufficientRes, node.Name)
 		}
 		if err := c.store.UpdateNodeResource(ctx, podname, node.Name, types.CPUMap{}, int64(len(containers))*memory, "-"); err != nil {
 			return err
@@ -259,7 +258,7 @@ func (c *Calcium) reallocNodesCPUMem(
 				}
 				// 这里只有1个节点，肯定会出现1个节点的解决方案
 				if total < need || len(nodeCPUPlans) != 1 {
-					return nil, fmt.Errorf("Not enough resource")
+					return nil, types.ErrInsufficientRes
 				}
 
 				cpuCost := types.CPUMap{}

cluster/calcium/resource.go

@@ -2,7 +2,6 @@ package calcium
 
 import (
 	"context"
-	"fmt"
 	"sort"
 
 	log "github.com/sirupsen/logrus"
@@ -45,7 +44,7 @@ func (c *Calcium) allocResource(ctx context.Context, opts *types.DeployOptions,
 	case scheduler.CPU_PRIOR:
 		nodesInfo, nodeCPUPlans, total, err = c.scheduler.SelectCPUNodes(nodesInfo, opts.CPUQuota, opts.Memory)
 	default:
-		return nil, fmt.Errorf("Pod type not support yet")
+		return nil, types.ErrBadPodType
 	}
 	if err != nil {
 		return nil, err
@@ -59,7 +58,7 @@ func (c *Calcium) allocResource(ctx context.Context, opts *types.DeployOptions,
 	case cluster.DeployFill:
 		nodesInfo, err = c.scheduler.FillDivision(nodesInfo, opts.Count, total)
 	default:
-		return nil, fmt.Errorf("Deploy method not support yet")
+		return nil, types.ErrBadDeployMethod
 	}
 	if err != nil {
 		return nil, err
@@ -70,7 +69,7 @@ func (c *Calcium) allocResource(ctx context.Context, opts *types.DeployOptions,
 	p := sort.Search(len(nodesInfo), func(i int) bool { return nodesInfo[i].Deploy > 0 })
 	// p 最大也就是 len(nodesInfo) - 1
 	if p == len(nodesInfo) {
-		return nil, fmt.Errorf("Cannot alloc a plan, not enough resource")
+		return nil, types.ErrInsufficientRes
 	}
 	nodesInfo = nodesInfo[p:]
 	for i, nodeInfo := range nodesInfo {
@@ -117,8 +116,7 @@ func (c *Calcium) getCPUAndMem(ctx context.Context, podname, nodename string, la
 	}
 
 	if len(nodes) == 0 {
-		err := fmt.Errorf("No available nodes")
-		return result, err
+		return result, types.ErrInsufficientNodes
 	}
 
 	result = makeCPUAndMem(nodes)

lock/etcdlock/lock.go

@@ -7,6 +7,7 @@ import (
 
 	"github.com/coreos/etcd/clientv3"
 	"github.com/coreos/etcd/clientv3/concurrency"
+	"github.com/projecteru2/core/types"
 	"golang.org/x/net/context"
 )
 
@@ -20,7 +21,7 @@ type Mutex struct {
 //New new a lock
 func New(cli *clientv3.Client, key string, ttl int) (*Mutex, error) {
 	if key == "" {
-		return nil, fmt.Errorf("No lock key")
+		return nil, types.ErrKeyIsEmpty
 	}
 
 	if !strings.HasPrefix(key, "/") {

network/sdn/plugin.go

@@ -18,12 +18,13 @@ type titanium struct{}
 // connect to network with ipv4 address
 func (t *titanium) ConnectToNetwork(ctx context.Context, containerID, networkID, ipv4 string) error {
 	if len(containerID) != 64 {
-		return fmt.Errorf("ContainerID must be in length of 64")
+		return types.ErrBadContainerID
 	}
 
 	engine, ok := utils.GetDockerEngineFromContext(ctx)
 	if !ok {
-		return fmt.Errorf("Not actually a `engineapi.Client` for value engine in context")
+		return types.NewDetailedErr(types.ErrCannotGetEngine,
+			fmt.Sprintf("Not actually a `engineapi.Client` for value engine in context"))
 	}
 
 	config := &enginenetwork.EndpointSettings{
@@ -35,7 +36,7 @@ func (t *titanium) ConnectToNetwork(ctx context.Context, containerID, networkID,
 	if ipv4 != "" {
 		ip := net.ParseIP(ipv4)
 		if ip == nil {
-			return fmt.Errorf("IP Address is not valid: %v", ipv4)
+			return types.NewDetailedErr(types.ErrBadIPAddress, ipv4)
 		}
 
 		config.IPAMConfig.IPv4Address = ip.String()
@@ -48,12 +49,13 @@ func (t *titanium) ConnectToNetwork(ctx context.Context, containerID, networkID,
 // disconnect from network
 func (t *titanium) DisconnectFromNetwork(ctx context.Context, containerID, networkID string) error {
 	if len(containerID) != 64 {
-		return fmt.Errorf("ContainerID must be in length of 64")
+		return types.ErrBadContainerID
 	}
 
 	engine, ok := utils.GetDockerEngineFromContext(ctx)
 	if !ok {
-		return fmt.Errorf("Not actually a `engineapi.Client` for value engine in context")
+		return types.NewDetailedErr(types.ErrCannotGetEngine,
+			fmt.Sprintf("Not actually a `engineapi.Client` for value engine in context"))
 	}
 
 	log.Debugf("[DisconnectFromNetwork] Disconnect %v from %v", containerID, networkID)
@@ -65,7 +67,8 @@ func (t *titanium) ListNetworks(ctx context.Context, driver string) ([]*types.Ne
 	networks := []*types.Network{}
 	engine, ok := utils.GetDockerEngineFromContext(ctx)
 	if !ok {
-		return networks, fmt.Errorf("Not actually a `engineapi.Client` for value engine in context")
+		return networks, types.NewDetailedErr(types.ErrCannotGetEngine,
+			fmt.Sprintf("Not actually a `engineapi.Client` for value engine in context"))
 	}
 
 	filters := enginefilters.NewArgs()

rpc/rpc_test.go

@@ -290,12 +290,12 @@ func TestContainers(t *testing.T) {
 	// test GetContainer error
 	store.On("GetContainer", "012").Return(&types.Container{}, nil)
 	_, err = clnt.GetContainer(ctx, &pb.ContainerID{Id: "012"})
-	assert.Contains(t, err.Error(), "Container ID must be length of 64")
+	assert.Contains(t, err.Error(), types.ErrBadContainerID.Error())
 
 	ID := "586f906185de3ed755d0db1c0f37149c1eae1ba557a26adccbe1f51c500d07d1"
 	store.On("GetContainer", ID).Return(&types.Container{}, nil)
 	_, err = clnt.GetContainer(ctx, &pb.ContainerID{Id: ID})
-	assert.Contains(t, err.Error(), "Engine is nil")
+	assert.Contains(t, err.Error(), types.ErrNilEngine.Error())
 
 	// test GetContainers
 	container := types.Container{
@@ -462,8 +462,7 @@ func TestCreateContainer(t *testing.T) {
 	resp, err := clnt.CreateContainer(ctx, &depOpts)
 	assert.NoError(t, err)
 	recv, err := resp.Recv()
-	assert.Error(t, err)
-	assert.Contains(t, err.Error(), "Minimum memory limit allowed is 4MB")
+	assert.Contains(t, err.Error(), types.ErrBadMemory.Error())
 	assert.Nil(t, recv)
 
 	depOpts = pb.DeployOptions{
@@ -487,8 +486,7 @@ func TestCreateContainer(t *testing.T) {
 	resp, err = clnt.CreateContainer(ctx, &depOpts)
 	assert.NoError(t, err)
 	recv, err = resp.Recv()
-	assert.Error(t, err)
-	assert.Contains(t, err.Error(), "Count must be positive")
+	assert.Contains(t, err.Error(), types.ErrBadCount.Error())
 	assert.Nil(t, recv)
 
 }
@@ -540,7 +538,7 @@ func TestRunAndWait(t *testing.T) {
 	assert.NoError(t, err)
 	assert.NoError(t, stream.Send(&runAndWaitOpts))
 	_, err = stream.Recv()
-	assert.Contains(t, err.Error(), "Minimum memory limit allowed is 4MB")
+	assert.Contains(t, err.Error(), types.ErrBadMemory.Error())
 
 	depOpts = pb.DeployOptions{
 		Image:     "",     // string
@@ -621,5 +619,5 @@ func TestOthers(t *testing.T) {
 	})
 	r, err := rmContainerResp.Recv() // 同理这个err也是gRPC调用的error，而不是执行动作的error
 	assert.Nil(t, err)
-	assert.Contains(t, r.GetMessage(), "Engine is nil")
+	assert.Contains(t, r.GetMessage(), types.ErrNilEngine.Error())
 }

scheduler/complex/average.go

@@ -1,7 +1,6 @@
 package complexscheduler
 
 import (
-	"fmt"
 	"sort"
 
 	"github.com/projecteru2/core/types"
@@ -14,7 +13,7 @@ func AveragePlan(nodesInfo []types.NodeInfo, need int) ([]types.NodeInfo, error)
 	sort.Slice(nodesInfo, func(i, j int) bool { return nodesInfo[i].Capacity < nodesInfo[j].Capacity })
 	p := sort.Search(nodesInfoLength, func(i int) bool { return nodesInfo[i].Capacity >= need })
 	if p == nodesInfoLength {
-		return nil, fmt.Errorf("Cannot alloc a each node plan, not enough capacity")
+		return nil, types.ErrInsufficientCap
 	}
 	nodesInfo = nodesInfo[p:]
 	for i := range nodesInfo {

scheduler/complex/cpu.go

@@ -5,7 +5,6 @@ CPU 分配的核心算法
 package complexscheduler
 
 import (
-	"fmt"
 	"sort"
 
 	"github.com/projecteru2/core/types"
@@ -196,7 +195,7 @@ func cpuPriorPlan(cpu float64, memory int64, nodesInfo []types.NodeInfo, maxShar
 	sort.Slice(nodesInfo, func(i, j int) bool { return nodesInfo[i].Capacity < nodesInfo[j].Capacity })
 	p := sort.Search(len(nodesInfo), func(i int) bool { return nodesInfo[i].Capacity > 0 })
 	if p == len(nodesInfo) {
-		return nil, nil, 0, fmt.Errorf("Not enough resource")
+		return nil, nil, 0, types.ErrInsufficientRes
 	}
 
 	log.Debugf("[cpuPriorPlan] nodesInfo: %v", nodesInfo)

scheduler/complex/fill.go

@@ -20,7 +20,8 @@ func FillPlan(nodesInfo []types.NodeInfo, need int) ([]types.NodeInfo, error) {
 	for i := range nodesInfo {
 		diff := need - nodesInfo[i].Count
 		if nodesInfo[i].Capacity < diff {
-			return nil, fmt.Errorf("Cannot alloc a fill node plan, node %s has not enough resource", nodesInfo[i].Name)
+			return nil, types.NewDetailedErr(types.ErrInsufficientRes,
+				fmt.Sprintf("node %s cannot alloc a fill node plan", nodesInfo[i].Name))
 		}
 		nodesInfo[i].Deploy = diff
 		nodesInfo[i].Capacity -= diff

scheduler/complex/fill_test.go

@@ -4,6 +4,7 @@ import (
 	"sort"
 	"testing"
 
+	"github.com/projecteru2/core/types"
 	"github.com/stretchr/testify/assert"
 )
 
@@ -49,8 +50,7 @@ func TestFillPlan(t *testing.T) {
 	n = 15
 	nodes = deployedNodes()
 	_, err = FillPlan(nodes, n)
-	assert.Error(t, err)
-	assert.Contains(t, err.Error(), "node n2 has not enough resource")
+	assert.EqualError(t, types.IsDetailedErr(err), types.ErrInsufficientRes.Error())
 
 	// 全局补充不能
 	n = 1