Currently ingress controller try to update status for each ingress resource in a parallel by using Goroutine, and inside this Goroutine we are trying to sort same IngressStatus reference which is shared between all Goroutine, this will break the original refrence if some Goroutine tried to sort exact same time. So we should have done sorting before passing reference to each Goroutine to prevent from breaking original reference fixes: #3269
398 lines
10 KiB
Go
398 lines
10 KiB
Go
/*
|
|
Copyright 2015 The Kubernetes Authors.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
*/
|
|
|
|
package status
|
|
|
|
import (
|
|
"fmt"
|
|
"net"
|
|
"os"
|
|
"sort"
|
|
"strings"
|
|
"time"
|
|
|
|
"github.com/golang/glog"
|
|
"github.com/pkg/errors"
|
|
|
|
pool "gopkg.in/go-playground/pool.v3"
|
|
apiv1 "k8s.io/api/core/v1"
|
|
extensions "k8s.io/api/extensions/v1beta1"
|
|
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
|
"k8s.io/apimachinery/pkg/labels"
|
|
"k8s.io/apimachinery/pkg/util/wait"
|
|
clientset "k8s.io/client-go/kubernetes"
|
|
"k8s.io/client-go/kubernetes/scheme"
|
|
"k8s.io/client-go/tools/leaderelection"
|
|
"k8s.io/client-go/tools/leaderelection/resourcelock"
|
|
"k8s.io/client-go/tools/record"
|
|
"k8s.io/kubernetes/pkg/kubelet/util/sliceutils"
|
|
|
|
"k8s.io/ingress-nginx/internal/k8s"
|
|
"k8s.io/ingress-nginx/internal/task"
|
|
)
|
|
|
|
const (
|
|
updateInterval = 60 * time.Second
|
|
)
|
|
|
|
// Sync ...
|
|
type Sync interface {
|
|
Run()
|
|
Shutdown()
|
|
}
|
|
|
|
type ingressLister interface {
|
|
// ListIngresses returns the list of Ingresses
|
|
ListIngresses() []*extensions.Ingress
|
|
}
|
|
|
|
// Config ...
|
|
type Config struct {
|
|
Client clientset.Interface
|
|
|
|
PublishService string
|
|
|
|
PublishStatusAddress string
|
|
|
|
ElectionID string
|
|
|
|
UpdateStatusOnShutdown bool
|
|
|
|
UseNodeInternalIP bool
|
|
|
|
IngressLister ingressLister
|
|
|
|
DefaultIngressClass string
|
|
IngressClass string
|
|
}
|
|
|
|
// statusSync keeps the status IP in each Ingress rule updated executing a periodic check
|
|
// in all the defined rules. To simplify the process leader election is used so the update
|
|
// is executed only in one node (Ingress controllers can be scaled to more than one)
|
|
// If the controller is running with the flag --publish-service (with a valid service)
|
|
// the IP address behind the service is used, if it is running with the flag
|
|
// --publish-status-address, the address specified in the flag is used, if neither of the
|
|
// two flags are set, the source is the IP/s of the node/s
|
|
type statusSync struct {
|
|
Config
|
|
// pod contains runtime information about this pod
|
|
pod *k8s.PodInfo
|
|
|
|
elector *leaderelection.LeaderElector
|
|
// workqueue used to keep in sync the status IP/s
|
|
// in the Ingress rules
|
|
syncQueue *task.Queue
|
|
}
|
|
|
|
// Run starts the loop to keep the status in sync
|
|
func (s statusSync) Run() {
|
|
s.elector.Run()
|
|
}
|
|
|
|
// Shutdown stop the sync. In case the instance is the leader it will remove the current IP
|
|
// if there is no other instances running.
|
|
func (s statusSync) Shutdown() {
|
|
go s.syncQueue.Shutdown()
|
|
// remove IP from Ingress
|
|
if !s.elector.IsLeader() {
|
|
return
|
|
}
|
|
|
|
if !s.UpdateStatusOnShutdown {
|
|
glog.Warningf("skipping update of status of Ingress rules")
|
|
return
|
|
}
|
|
|
|
glog.Infof("updating status of Ingress rules (remove)")
|
|
|
|
addrs, err := s.runningAddresses()
|
|
if err != nil {
|
|
glog.Errorf("error obtaining running IPs: %v", addrs)
|
|
return
|
|
}
|
|
|
|
if len(addrs) > 1 {
|
|
// leave the job to the next leader
|
|
glog.Infof("leaving status update for next leader (%v)", len(addrs))
|
|
return
|
|
}
|
|
|
|
if s.isRunningMultiplePods() {
|
|
glog.V(2).Infof("skipping Ingress status update (multiple pods running - another one will be elected as master)")
|
|
return
|
|
}
|
|
|
|
glog.Infof("removing address from ingress status (%v)", addrs)
|
|
s.updateStatus([]apiv1.LoadBalancerIngress{})
|
|
}
|
|
|
|
func (s *statusSync) sync(key interface{}) error {
|
|
if s.syncQueue.IsShuttingDown() {
|
|
glog.V(2).Infof("skipping Ingress status update (shutting down in progress)")
|
|
return nil
|
|
}
|
|
|
|
addrs, err := s.runningAddresses()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
s.updateStatus(sliceToStatus(addrs))
|
|
|
|
return nil
|
|
}
|
|
|
|
func (s statusSync) keyfunc(input interface{}) (interface{}, error) {
|
|
return input, nil
|
|
}
|
|
|
|
// NewStatusSyncer returns a new Sync instance
|
|
func NewStatusSyncer(config Config) Sync {
|
|
pod, err := k8s.GetPodDetails(config.Client)
|
|
if err != nil {
|
|
glog.Fatalf("unexpected error obtaining pod information: %v", err)
|
|
}
|
|
|
|
st := statusSync{
|
|
pod: pod,
|
|
|
|
Config: config,
|
|
}
|
|
st.syncQueue = task.NewCustomTaskQueue(st.sync, st.keyfunc)
|
|
|
|
// we need to use the defined ingress class to allow multiple leaders
|
|
// in order to update information about ingress status
|
|
electionID := fmt.Sprintf("%v-%v", config.ElectionID, config.DefaultIngressClass)
|
|
if config.IngressClass != "" {
|
|
electionID = fmt.Sprintf("%v-%v", config.ElectionID, config.IngressClass)
|
|
}
|
|
|
|
callbacks := leaderelection.LeaderCallbacks{
|
|
OnStartedLeading: func(stop <-chan struct{}) {
|
|
glog.V(2).Infof("I am the new status update leader")
|
|
go st.syncQueue.Run(time.Second, stop)
|
|
// when this instance is the leader we need to enqueue
|
|
// an item to trigger the update of the Ingress status.
|
|
wait.PollUntil(updateInterval, func() (bool, error) {
|
|
st.syncQueue.EnqueueTask(task.GetDummyObject("sync status"))
|
|
return false, nil
|
|
}, stop)
|
|
},
|
|
OnStoppedLeading: func() {
|
|
glog.V(2).Infof("I am not status update leader anymore")
|
|
},
|
|
OnNewLeader: func(identity string) {
|
|
glog.Infof("new leader elected: %v", identity)
|
|
},
|
|
}
|
|
|
|
broadcaster := record.NewBroadcaster()
|
|
hostname, _ := os.Hostname()
|
|
|
|
recorder := broadcaster.NewRecorder(scheme.Scheme, apiv1.EventSource{
|
|
Component: "ingress-leader-elector",
|
|
Host: hostname,
|
|
})
|
|
|
|
lock := resourcelock.ConfigMapLock{
|
|
ConfigMapMeta: metav1.ObjectMeta{Namespace: pod.Namespace, Name: electionID},
|
|
Client: config.Client.CoreV1(),
|
|
LockConfig: resourcelock.ResourceLockConfig{
|
|
Identity: pod.Name,
|
|
EventRecorder: recorder,
|
|
},
|
|
}
|
|
|
|
ttl := 30 * time.Second
|
|
le, err := leaderelection.NewLeaderElector(leaderelection.LeaderElectionConfig{
|
|
Lock: &lock,
|
|
LeaseDuration: ttl,
|
|
RenewDeadline: ttl / 2,
|
|
RetryPeriod: ttl / 4,
|
|
Callbacks: callbacks,
|
|
})
|
|
|
|
if err != nil {
|
|
glog.Fatalf("unexpected error starting leader election: %v", err)
|
|
}
|
|
|
|
st.elector = le
|
|
return st
|
|
}
|
|
|
|
// runningAddresses returns a list of IP addresses and/or FQDN where the
|
|
// ingress controller is currently running
|
|
func (s *statusSync) runningAddresses() ([]string, error) {
|
|
addrs := []string{}
|
|
|
|
if s.PublishService != "" {
|
|
ns, name, _ := k8s.ParseNameNS(s.PublishService)
|
|
svc, err := s.Client.CoreV1().Services(ns).Get(name, metav1.GetOptions{})
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
if svc.Spec.Type == apiv1.ServiceTypeExternalName {
|
|
addrs = append(addrs, svc.Spec.ExternalName)
|
|
return addrs, nil
|
|
}
|
|
|
|
for _, ip := range svc.Status.LoadBalancer.Ingress {
|
|
if ip.IP == "" {
|
|
addrs = append(addrs, ip.Hostname)
|
|
} else {
|
|
addrs = append(addrs, ip.IP)
|
|
}
|
|
}
|
|
|
|
addrs = append(addrs, svc.Spec.ExternalIPs...)
|
|
return addrs, nil
|
|
}
|
|
|
|
if s.PublishStatusAddress != "" {
|
|
addrs = append(addrs, s.PublishStatusAddress)
|
|
return addrs, nil
|
|
}
|
|
|
|
// get information about all the pods running the ingress controller
|
|
pods, err := s.Client.CoreV1().Pods(s.pod.Namespace).List(metav1.ListOptions{
|
|
LabelSelector: labels.SelectorFromSet(s.pod.Labels).String(),
|
|
})
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
for _, pod := range pods.Items {
|
|
// only Running pods are valid
|
|
if pod.Status.Phase != apiv1.PodRunning {
|
|
continue
|
|
}
|
|
|
|
name := k8s.GetNodeIPOrName(s.Client, pod.Spec.NodeName, s.UseNodeInternalIP)
|
|
if !sliceutils.StringInSlice(name, addrs) {
|
|
addrs = append(addrs, name)
|
|
}
|
|
}
|
|
|
|
return addrs, nil
|
|
}
|
|
|
|
func (s *statusSync) isRunningMultiplePods() bool {
|
|
pods, err := s.Client.CoreV1().Pods(s.pod.Namespace).List(metav1.ListOptions{
|
|
LabelSelector: labels.SelectorFromSet(s.pod.Labels).String(),
|
|
})
|
|
if err != nil {
|
|
return false
|
|
}
|
|
|
|
return len(pods.Items) > 1
|
|
}
|
|
|
|
// sliceToStatus converts a slice of IP and/or hostnames to LoadBalancerIngress
|
|
func sliceToStatus(endpoints []string) []apiv1.LoadBalancerIngress {
|
|
lbi := []apiv1.LoadBalancerIngress{}
|
|
for _, ep := range endpoints {
|
|
if net.ParseIP(ep) == nil {
|
|
lbi = append(lbi, apiv1.LoadBalancerIngress{Hostname: ep})
|
|
} else {
|
|
lbi = append(lbi, apiv1.LoadBalancerIngress{IP: ep})
|
|
}
|
|
}
|
|
|
|
sort.SliceStable(lbi, func(a, b int) bool {
|
|
return lbi[a].IP < lbi[b].IP
|
|
})
|
|
|
|
return lbi
|
|
}
|
|
|
|
// updateStatus changes the status information of Ingress rules
|
|
func (s *statusSync) updateStatus(newIngressPoint []apiv1.LoadBalancerIngress) {
|
|
ings := s.IngressLister.ListIngresses()
|
|
|
|
p := pool.NewLimited(10)
|
|
defer p.Close()
|
|
|
|
batch := p.Batch()
|
|
sort.SliceStable(newIngressPoint, lessLoadBalancerIngress(newIngressPoint))
|
|
|
|
for _, ing := range ings {
|
|
batch.Queue(runUpdate(ing, newIngressPoint, s.Client))
|
|
}
|
|
|
|
batch.QueueComplete()
|
|
batch.WaitAll()
|
|
}
|
|
|
|
func runUpdate(ing *extensions.Ingress, status []apiv1.LoadBalancerIngress,
|
|
client clientset.Interface) pool.WorkFunc {
|
|
return func(wu pool.WorkUnit) (interface{}, error) {
|
|
if wu.IsCancelled() {
|
|
return nil, nil
|
|
}
|
|
|
|
curIPs := ing.Status.LoadBalancer.Ingress
|
|
sort.SliceStable(curIPs, lessLoadBalancerIngress(curIPs))
|
|
|
|
if ingressSliceEqual(status, curIPs) {
|
|
glog.V(3).Infof("skipping update of Ingress %v/%v (no change)", ing.Namespace, ing.Name)
|
|
return true, nil
|
|
}
|
|
|
|
ingClient := client.ExtensionsV1beta1().Ingresses(ing.Namespace)
|
|
|
|
currIng, err := ingClient.Get(ing.Name, metav1.GetOptions{})
|
|
if err != nil {
|
|
return nil, errors.Wrap(err, fmt.Sprintf("unexpected error searching Ingress %v/%v", ing.Namespace, ing.Name))
|
|
}
|
|
|
|
glog.Infof("updating Ingress %v/%v status to %v", currIng.Namespace, currIng.Name, status)
|
|
currIng.Status.LoadBalancer.Ingress = status
|
|
_, err = ingClient.UpdateStatus(currIng)
|
|
if err != nil {
|
|
glog.Warningf("error updating ingress rule: %v", err)
|
|
}
|
|
|
|
return true, nil
|
|
}
|
|
}
|
|
|
|
func lessLoadBalancerIngress(addrs []apiv1.LoadBalancerIngress) func(int, int) bool {
|
|
return func(a, b int) bool {
|
|
switch strings.Compare(addrs[a].Hostname, addrs[b].Hostname) {
|
|
case -1:
|
|
return true
|
|
case 1:
|
|
return false
|
|
}
|
|
return addrs[a].IP < addrs[b].IP
|
|
}
|
|
}
|
|
|
|
func ingressSliceEqual(lhs, rhs []apiv1.LoadBalancerIngress) bool {
|
|
if len(lhs) != len(rhs) {
|
|
return false
|
|
}
|
|
|
|
for i := range lhs {
|
|
if lhs[i].IP != rhs[i].IP {
|
|
return false
|
|
}
|
|
if lhs[i].Hostname != rhs[i].Hostname {
|
|
return false
|
|
}
|
|
}
|
|
return true
|
|
}
|