Rainbond/eventlog/cluster/discover/manager.go
2020-07-23 11:26:16 +08:00

411 lines
12 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

// Copyright (C) 2014-2018 Goodrain Co., Ltd.
// RAINBOND, Application Management Platform
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version. For any non-GPL usage of Rainbond,
// one or multiple Commercial Licenses authorized by Goodrain Co., Ltd.
// must be obtained first.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
// You should have received a copy of the GNU General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package discover
import (
"encoding/json"
"fmt"
"net"
"os"
"strings"
"time"
"github.com/Sirupsen/logrus"
"github.com/coreos/etcd/client"
"github.com/coreos/etcd/clientv3"
"github.com/coreos/etcd/mvcc/mvccpb"
"github.com/goodrain/rainbond/eventlog/conf"
"github.com/goodrain/rainbond/eventlog/util"
etcdutil "github.com/goodrain/rainbond/util/etcd"
"github.com/prometheus/client_golang/prometheus"
"golang.org/x/net/context"
)
//Manager 节点动态发现管理器
type Manager interface {
RegisteredInstance(host string, port int, stopRegister *bool) *Instance
CancellationInstance(instance *Instance)
MonitorAddInstances() chan *Instance
MonitorDelInstances() chan *Instance
MonitorUpdateInstances() chan *Instance
GetInstance(string) *Instance
InstanceCheckHealth(string) string
Run() error
GetCurrentInstance() Instance
Stop()
Scrape(ch chan<- prometheus.Metric, namespace, exporter string) error
}
//EtcdDiscoverManager 基于ETCD自动发现
type EtcdDiscoverManager struct {
cancel func()
context context.Context
addChan chan *Instance
delChan chan *Instance
updateChan chan *Instance
log *logrus.Entry
conf conf.DiscoverConf
etcdclientv3 *clientv3.Client
selfInstance *Instance
othersInstance []*Instance
stopDiscover bool
}
//New 创建
func New(etcdClient *clientv3.Client, conf conf.DiscoverConf, log *logrus.Entry) Manager {
ctx, cancel := context.WithCancel(context.Background())
return &EtcdDiscoverManager{
conf: conf,
cancel: cancel,
context: ctx,
log: log,
addChan: make(chan *Instance, 2),
delChan: make(chan *Instance, 2),
updateChan: make(chan *Instance, 2),
othersInstance: make([]*Instance, 0),
etcdclientv3: etcdClient,
}
}
//GetCurrentInstance 获取当前节点
func (d *EtcdDiscoverManager) GetCurrentInstance() Instance {
return *d.selfInstance
}
//RegisteredInstance 注册实例
func (d *EtcdDiscoverManager) RegisteredInstance(host string, port int, stopRegister *bool) *Instance {
instance := &Instance{}
for !*stopRegister {
if host == "0.0.0.0" || host == "127.0.0.1" || host == "localhost" {
if d.conf.InstanceIP != "" {
ip := net.ParseIP(d.conf.InstanceIP)
if ip != nil {
instance.HostIP = ip
}
}
} else {
ip := net.ParseIP(host)
if ip != nil {
instance.HostIP = ip
}
}
if instance.HostIP == nil {
ip, err := util.ExternalIP()
if err != nil {
d.log.Error("Can not get host ip for the instance.")
time.Sleep(time.Second * 10)
continue
} else {
instance.HostIP = ip
}
}
instance.PubPort = port
instance.DockerLogPort = d.conf.DockerLogPort
instance.WebPort = d.conf.WebPort
instance.HostID = d.conf.NodeID
instance.HostName, _ = os.Hostname()
instance.Status = "create"
data, err := json.Marshal(instance)
if err != nil {
d.log.Error("Create register instance data error.", err.Error())
time.Sleep(time.Second * 10)
continue
}
ctx, cancel := context.WithCancel(d.context)
_, err = d.etcdclientv3.Put(ctx, fmt.Sprintf("%s/instance/%s:%d", d.conf.HomePath, instance.HostIP, instance.PubPort), string(data))
cancel()
if err != nil {
d.log.Error("Register instance data to etcd error.", err.Error())
time.Sleep(time.Second * 10)
continue
}
d.selfInstance = instance
go d.discover()
d.log.Infof("Register instance in cluster success. HostID:%s HostIP:%s PubPort:%d", instance.HostID, instance.HostIP, instance.PubPort)
return instance
}
return nil
}
//MonitorAddInstances 实例通知
func (d *EtcdDiscoverManager) MonitorAddInstances() chan *Instance {
return d.addChan
}
//MonitorDelInstances 实例通知
func (d *EtcdDiscoverManager) MonitorDelInstances() chan *Instance {
return d.delChan
}
//MonitorUpdateInstances 实例通知
func (d *EtcdDiscoverManager) MonitorUpdateInstances() chan *Instance {
return d.updateChan
}
//Run 启动
func (d *EtcdDiscoverManager) Run() error {
d.log.Info("Discover manager start ")
etcdClientArgs := &etcdutil.ClientArgs{
Endpoints: d.conf.EtcdAddr,
CaFile: d.conf.EtcdCaFile,
CertFile: d.conf.EtcdCertFile,
KeyFile: d.conf.EtcdKeyFile,
}
var err error
d.etcdclientv3, err = etcdutil.NewClient(d.context, etcdClientArgs)
if err != nil {
d.log.Error("Create etcd v3 client error.", err.Error())
return err
}
return nil
}
//Discover 发现
func (d *EtcdDiscoverManager) discover() {
tike := time.NewTicker(time.Second * 5)
defer tike.Stop()
for {
ctx, cancel := context.WithCancel(d.context)
res, err := d.etcdclientv3.Get(ctx, d.conf.HomePath+"/instance/", clientv3.WithPrefix())
cancel()
if err != nil {
d.log.Error("Get instance info from etcd error.", err.Error())
} else {
for _, kv := range res.Kvs {
node := &client.Node{
Key: string(kv.Key),
Value: string(kv.Value),
}
d.add(node)
}
break
}
select {
case <-tike.C:
case <-d.context.Done():
return
}
}
ctx, cancel := context.WithCancel(d.context)
defer cancel()
watcher := d.etcdclientv3.Watch(ctx, d.conf.HomePath+"/instance/", clientv3.WithPrefix())
for !d.stopDiscover {
res, ok := <-watcher
if !ok {
break
}
for _, event := range res.Events {
node := &client.Node{
Key: string(event.Kv.Key),
Value: string(event.Kv.Value),
}
switch event.Type {
case mvccpb.PUT:
d.add(node)
case mvccpb.DELETE:
//忽略自己
if strings.HasSuffix(node.Key, fmt.Sprintf("/%s:%d", d.selfInstance.HostIP, d.selfInstance.PubPort)) {
continue
}
keys := strings.Split(node.Key, "/")
hostPort := keys[len(keys)-1]
d.log.Infof("Delete an instance.%s", hostPort)
var removeIndex int
var have bool
for i, ins := range d.othersInstance {
if fmt.Sprintf("%s:%d", ins.HostIP, ins.PubPort) == hostPort {
removeIndex = i
have = true
break
}
}
if have {
instance := d.othersInstance[removeIndex]
d.othersInstance = DeleteSlice(d.othersInstance, removeIndex)
d.MonitorDelInstances() <- instance
d.log.Infof("A instance offline %s", instance.HostName)
}
}
}
}
d.log.Debug("discover manager discover core stop")
}
func (d *EtcdDiscoverManager) add(node *client.Node) {
//忽略自己
if strings.HasSuffix(node.Key, fmt.Sprintf("/%s:%d", d.selfInstance.HostIP, d.selfInstance.PubPort)) {
return
}
var instance Instance
if err := json.Unmarshal([]byte(node.Value), &instance); err != nil {
d.log.Error("Unmarshal instance data that from etcd error.", err.Error())
} else {
if strings.HasSuffix(node.Key, fmt.Sprintf("/%s:%d", d.selfInstance.HostIP, d.selfInstance.PubPort)) {
d.selfInstance = &instance
}
isExist := false
for _, i := range d.othersInstance {
if i.HostID == instance.HostID {
*i = instance
d.log.Debug("update the instance " + i.HostID)
isExist = true
break
}
}
if !isExist {
d.log.Infof("Find an instance.IP:%s, Port:%d, NodeID:%s HostID: %s", instance.HostIP.String(), instance.PubPort, instance.HostName, instance.HostID)
d.MonitorAddInstances() <- &instance
d.othersInstance = append(d.othersInstance, &instance)
}
}
}
func (d *EtcdDiscoverManager) update(node *client.Node) {
var instance Instance
if err := json.Unmarshal([]byte(node.Value), &instance); err != nil {
d.log.Error("Unmarshal instance data that from etcd error.", err.Error())
} else {
if strings.HasSuffix(node.Key, fmt.Sprintf("/%s:%d", d.selfInstance.HostIP, d.selfInstance.PubPort)) {
d.selfInstance = &instance
}
for _, i := range d.othersInstance {
if i.HostID == instance.HostID {
*i = instance
d.log.Debug("update the instance " + i.HostID)
}
}
}
}
//DeleteSlice 从数组中删除某元素
func DeleteSlice(source []*Instance, index int) []*Instance {
if len(source) == 1 {
return make([]*Instance, 0)
}
if index == 0 {
return source[1:]
}
if index == len(source)-1 {
return source[:len(source)-2]
}
return append(source[0:index-1], source[index+1:]...)
}
//Stop 停止
func (d *EtcdDiscoverManager) Stop() {
d.stopDiscover = true
d.cancel()
d.log.Info("Stop the discover manager.")
}
//CancellationInstance 注销实例
func (d *EtcdDiscoverManager) CancellationInstance(instance *Instance) {
ctx, cancel := context.WithTimeout(d.context, time.Second*5)
defer cancel()
_, err := d.etcdclientv3.Delete(ctx, fmt.Sprintf("%s/instance/%s:%d", d.conf.HomePath, instance.HostIP, instance.PubPort))
if err != nil && !client.IsKeyNotFound(err) {
d.log.Error("Cancellation Instance from etcd error.", err.Error())
} else {
d.log.Info("Cancellation Instance from etcd")
}
}
//UpdateInstance 更新实例
func (d *EtcdDiscoverManager) UpdateInstance(instance *Instance) {
instance.Status = "update"
data, err := json.Marshal(instance)
if err != nil {
d.log.Error("Create update instance data error.", err.Error())
return
}
ctx, cancel := context.WithTimeout(d.context, time.Second*5)
defer cancel()
_, err = d.etcdclientv3.Put(ctx, fmt.Sprintf("%s/instance/%s:%d", d.conf.HomePath, instance.HostIP, instance.PubPort), string(data))
if err != nil && !client.IsKeyNotFound(err) {
d.log.Error(" Update Instance from etcd error.", err.Error())
}
}
//InstanceCheckHealth 将由distribution调用当发现节点不正常时
//此处检查,如果节点已经下线,返回 delete
//如果节点未下线标记为异常,返回 abnormal
//如果节点被集群判断为故障,返回 delete
func (d *EtcdDiscoverManager) InstanceCheckHealth(instanceID string) string {
d.log.Info("Start check instance health.")
if d.selfInstance.HostID == instanceID {
d.log.Error("The current node condition monitoring.")
return "abnormal"
}
for _, i := range d.othersInstance {
if i.HostID == instanceID {
d.log.Errorf("Instance (%s) is abnormal.", instanceID)
i.Status = "abnormal"
i.TagNumber++
if i.TagNumber > ((len(d.othersInstance) + 1) / 2) { //大于一半的节点标记
d.log.Warn("Instance (%s) is abnormal. tag number more than half of all instance number. will cancellation.", instanceID)
d.CancellationInstance(i)
return "delete"
}
d.UpdateInstance(i)
return "abnormal"
}
}
return "delete"
}
//GetInstance 获取实例
func (d *EtcdDiscoverManager) GetInstance(id string) *Instance {
if id == d.selfInstance.HostID {
return d.selfInstance
}
for _, i := range d.othersInstance {
if i.HostID == id {
return i
}
}
return nil
}
//Scrape prometheus monitor metrics
func (d *EtcdDiscoverManager) Scrape(ch chan<- prometheus.Metric, namespace, exporter string) error {
instanceDesc := prometheus.NewDesc(
prometheus.BuildFQName(namespace, exporter, "instance_up"),
"the instance in cluster status.",
[]string{"from", "instance", "status"}, nil,
)
for _, i := range d.othersInstance {
if i.Status == "delete" || i.Status == "abnormal" {
ch <- prometheus.MustNewConstMetric(instanceDesc, prometheus.GaugeValue, 0, d.selfInstance.HostIP.String(), i.HostIP.String(), i.Status)
} else {
ch <- prometheus.MustNewConstMetric(instanceDesc, prometheus.GaugeValue, 1, d.selfInstance.HostIP.String(), i.HostIP.String(), i.Status)
}
}
return nil
}