// Because openshift-sdn uses an overlay and doesn't need GCE Routes, we need to
// clear the NetworkUnavailable condition that kubelet adds to initial node
// status when using GCE.
// TODO: make upstream kubelet more flexible with overlays and GCE so this
// condition doesn't get added for network plugins that don't want it, and then
// we can remove this function.
func (master *OsdnMaster) clearInitialNodeNetworkUnavailableCondition(node *kapi.Node) {
knode := node
cleared := false
resultErr := kclient.RetryOnConflict(kclient.DefaultBackoff, func() error {
var err error
if knode != node {
knode, err = master.kClient.Nodes().Get(node.ObjectMeta.Name)
if err != nil {
return err
}
}
// Let caller modify knode's status, then push to api server.
_, condition := kapi.GetNodeCondition(&node.Status, kapi.NodeNetworkUnavailable)
if condition != nil && condition.Status != kapi.ConditionFalse && condition.Reason == "NoRouteCreated" {
condition.Status = kapi.ConditionFalse
condition.Reason = "RouteCreated"
condition.Message = "openshift-sdn cleared kubelet-set NoRouteCreated"
condition.LastTransitionTime = kapiunversioned.Now()
knode, err = master.kClient.Nodes().UpdateStatus(knode)
if err == nil {
cleared = true
}
}
return err
})
if resultErr != nil {
utilruntime.HandleError(fmt.Errorf("Status update failed for local node: %v", resultErr))
} else if cleared {
log.Infof("Cleared node NetworkUnavailable/NoRouteCreated condition for %s", node.ObjectMeta.Name)
}
}
func updateNetworkingCondition(node *api.Node, routeCreated bool) {
_, networkingCondition := api.GetNodeCondition(&node.Status, api.NodeNetworkUnavailable)
currentTime := unversioned.Now()
if routeCreated {
if networkingCondition != nil && networkingCondition.Status != api.ConditionFalse {
networkingCondition.Status = api.ConditionFalse
networkingCondition.Reason = "RouteCreated"
networkingCondition.Message = "RouteController created a route"
networkingCondition.LastTransitionTime = currentTime
} else if networkingCondition == nil {
node.Status.Conditions = append(node.Status.Conditions, api.NodeCondition{
Type: api.NodeNetworkUnavailable,
Status: api.ConditionFalse,
Reason: "RouteCreated",
Message: "RouteController created a route",
LastTransitionTime: currentTime,
})
}
} else {
if networkingCondition != nil && networkingCondition.Status != api.ConditionTrue {
networkingCondition.Status = api.ConditionTrue
networkingCondition.Reason = "NoRouteCreated"
networkingCondition.Message = "RouteController failed to create a route"
networkingCondition.LastTransitionTime = currentTime
} else if networkingCondition == nil {
node.Status.Conditions = append(node.Status.Conditions, api.NodeCondition{
Type: api.NodeNetworkUnavailable,
Status: api.ConditionTrue,
Reason: "NoRouteCreated",
Message: "RouteController failed to create a route",
LastTransitionTime: currentTime,
})
}
}
}
// verifyCondition verifies specific node condition is generated, if reason and message are empty, they will not be checked
func verifyCondition(n client.NodeInterface, nodeName string, condition api.NodeConditionType, status api.ConditionStatus, reason, message string) error {
node, err := n.Get(nodeName)
if err != nil {
return err
}
_, c := api.GetNodeCondition(&node.Status, condition)
if c == nil {
return fmt.Errorf("node condition %q not found", condition)
}
if c.Status != status || c.Reason != reason || c.Message != message {
return fmt.Errorf("unexpected node condition %q: %+v", condition, c)
}
return nil
}
// For a given node checks its conditions and tries to update it. Returns grace period to which given node
// is entitled, state of current and last observed Ready Condition, and an error if it occurred.
func (nc *NodeController) tryUpdateNodeStatus(node *api.Node) (time.Duration, api.NodeCondition, *api.NodeCondition, error) {
var err error
var gracePeriod time.Duration
var observedReadyCondition api.NodeCondition
_, currentReadyCondition := api.GetNodeCondition(&node.Status, api.NodeReady)
if currentReadyCondition == nil {
// If ready condition is nil, then kubelet (or nodecontroller) never posted node status.
// A fake ready condition is created, where LastProbeTime and LastTransitionTime is set
// to node.CreationTimestamp to avoid handle the corner case.
observedReadyCondition = api.NodeCondition{
Type: api.NodeReady,
Status: api.ConditionUnknown,
LastHeartbeatTime: node.CreationTimestamp,
LastTransitionTime: node.CreationTimestamp,
}
gracePeriod = nc.nodeStartupGracePeriod
nc.nodeStatusMap[node.Name] = nodeStatusData{
status: node.Status,
probeTimestamp: node.CreationTimestamp,
readyTransitionTimestamp: node.CreationTimestamp,
}
} else {
// If ready condition is not nil, make a copy of it, since we may modify it in place later.
observedReadyCondition = *currentReadyCondition
gracePeriod = nc.nodeMonitorGracePeriod
}
savedNodeStatus, found := nc.nodeStatusMap[node.Name]
// There are following cases to check:
// - both saved and new status have no Ready Condition set - we leave everything as it is,
// - saved status have no Ready Condition, but current one does - NodeController was restarted with Node data already present in etcd,
// - saved status have some Ready Condition, but current one does not - it's an error, but we fill it up because that's probably a good thing to do,
// - both saved and current statuses have Ready Conditions and they have the same LastProbeTime - nothing happened on that Node, it may be
// unresponsive, so we leave it as it is,
// - both saved and current statuses have Ready Conditions, they have different LastProbeTimes, but the same Ready Condition State -
// everything's in order, no transition occurred, we update only probeTimestamp,
// - both saved and current statuses have Ready Conditions, different LastProbeTimes and different Ready Condition State -
// Ready Condition changed it state since we last seen it, so we update both probeTimestamp and readyTransitionTimestamp.
// TODO: things to consider:
// - if 'LastProbeTime' have gone back in time its probably an error, currently we ignore it,
// - currently only correct Ready State transition outside of Node Controller is marking it ready by Kubelet, we don't check
// if that's the case, but it does not seem necessary.
var savedCondition *api.NodeCondition
if found {
_, savedCondition = api.GetNodeCondition(&savedNodeStatus.status, api.NodeReady)
}
_, observedCondition := api.GetNodeCondition(&node.Status, api.NodeReady)
if !found {
glog.Warningf("Missing timestamp for Node %s. Assuming now as a timestamp.", node.Name)
savedNodeStatus = nodeStatusData{
status: node.Status,
probeTimestamp: nc.now(),
readyTransitionTimestamp: nc.now(),
}
} else if savedCondition == nil && observedCondition != nil {
glog.V(1).Infof("Creating timestamp entry for newly observed Node %s", node.Name)
savedNodeStatus = nodeStatusData{
status: node.Status,
probeTimestamp: nc.now(),
readyTransitionTimestamp: nc.now(),
}
} else if savedCondition != nil && observedCondition == nil {
glog.Errorf("ReadyCondition was removed from Status of Node %s", node.Name)
// TODO: figure out what to do in this case. For now we do the same thing as above.
savedNodeStatus = nodeStatusData{
status: node.Status,
probeTimestamp: nc.now(),
readyTransitionTimestamp: nc.now(),
}
} else if savedCondition != nil && observedCondition != nil && savedCondition.LastHeartbeatTime != observedCondition.LastHeartbeatTime {
var transitionTime unversioned.Time
// If ReadyCondition changed since the last time we checked, we update the transition timestamp to "now",
// otherwise we leave it as it is.
if savedCondition.LastTransitionTime != observedCondition.LastTransitionTime {
glog.V(3).Infof("ReadyCondition for Node %s transitioned from %v to %v", node.Name, savedCondition.Status, observedCondition)
transitionTime = nc.now()
} else {
transitionTime = savedNodeStatus.readyTransitionTimestamp
}
if glog.V(5) {
glog.V(5).Infof("Node %s ReadyCondition updated. Updating timestamp: %+v vs %+v.", node.Name, savedNodeStatus.status, node.Status)
} else {
glog.V(3).Infof("Node %s ReadyCondition updated. Updating timestamp.", node.Name)
}
savedNodeStatus = nodeStatusData{
status: node.Status,
probeTimestamp: nc.now(),
readyTransitionTimestamp: transitionTime,
}
}
nc.nodeStatusMap[node.Name] = savedNodeStatus
if nc.now().After(savedNodeStatus.probeTimestamp.Add(gracePeriod)) {
// NodeReady condition was last set longer ago than gracePeriod, so update it to Unknown
// (regardless of its current value) in the master.
if currentReadyCondition == nil {
glog.V(2).Infof("node %v is never updated by kubelet", node.Name)
node.Status.Conditions = append(node.Status.Conditions, api.NodeCondition{
Type: api.NodeReady,
Status: api.ConditionUnknown,
Reason: "NodeStatusNeverUpdated",
Message: fmt.Sprintf("Kubelet never posted node status."),
LastHeartbeatTime: node.CreationTimestamp,
LastTransitionTime: nc.now(),
})
} else {
glog.V(4).Infof("node %v hasn't been updated for %+v. Last ready condition is: %+v",
node.Name, nc.now().Time.Sub(savedNodeStatus.probeTimestamp.Time), observedReadyCondition)
if observedReadyCondition.Status != api.ConditionUnknown {
currentReadyCondition.Status = api.ConditionUnknown
currentReadyCondition.Reason = "NodeStatusUnknown"
currentReadyCondition.Message = fmt.Sprintf("Kubelet stopped posting node status.")
// LastProbeTime is the last time we heard from kubelet.
currentReadyCondition.LastHeartbeatTime = observedReadyCondition.LastHeartbeatTime
currentReadyCondition.LastTransitionTime = nc.now()
}
}
// Like NodeReady condition, NodeOutOfDisk was last set longer ago than gracePeriod, so update
// it to Unknown (regardless of its current value) in the master.
// TODO(madhusudancs): Refactor this with readyCondition to remove duplicated code.
_, oodCondition := api.GetNodeCondition(&node.Status, api.NodeOutOfDisk)
if oodCondition == nil {
glog.V(2).Infof("Out of disk condition of node %v is never updated by kubelet", node.Name)
node.Status.Conditions = append(node.Status.Conditions, api.NodeCondition{
Type: api.NodeOutOfDisk,
Status: api.ConditionUnknown,
Reason: "NodeStatusNeverUpdated",
Message: fmt.Sprintf("Kubelet never posted node status."),
LastHeartbeatTime: node.CreationTimestamp,
LastTransitionTime: nc.now(),
})
} else {
glog.V(4).Infof("node %v hasn't been updated for %+v. Last out of disk condition is: %+v",
node.Name, nc.now().Time.Sub(savedNodeStatus.probeTimestamp.Time), oodCondition)
if oodCondition.Status != api.ConditionUnknown {
oodCondition.Status = api.ConditionUnknown
oodCondition.Reason = "NodeStatusUnknown"
oodCondition.Message = fmt.Sprintf("Kubelet stopped posting node status.")
oodCondition.LastTransitionTime = nc.now()
}
}
_, currentCondition := api.GetNodeCondition(&node.Status, api.NodeReady)
if !api.Semantic.DeepEqual(currentCondition, &observedReadyCondition) {
if _, err = nc.kubeClient.Core().Nodes().UpdateStatus(node); err != nil {
glog.Errorf("Error updating node %s: %v", node.Name, err)
return gracePeriod, observedReadyCondition, currentReadyCondition, err
} else {
nc.nodeStatusMap[node.Name] = nodeStatusData{
status: node.Status,
probeTimestamp: nc.nodeStatusMap[node.Name].probeTimestamp,
readyTransitionTimestamp: nc.now(),
}
return gracePeriod, observedReadyCondition, currentReadyCondition, nil
}
}
}
return gracePeriod, observedReadyCondition, currentReadyCondition, err
}
func TestReconcile(t *testing.T) {
cluster := "my-k8s"
node1 := api.Node{ObjectMeta: api.ObjectMeta{Name: "node-1", UID: "01"}, Spec: api.NodeSpec{PodCIDR: "10.120.0.0/24"}}
node2 := api.Node{ObjectMeta: api.ObjectMeta{Name: "node-2", UID: "02"}, Spec: api.NodeSpec{PodCIDR: "10.120.1.0/24"}}
nodeNoCidr := api.Node{ObjectMeta: api.ObjectMeta{Name: "node-2", UID: "02"}, Spec: api.NodeSpec{PodCIDR: ""}}
testCases := []struct {
nodes []api.Node
initialRoutes []*cloudprovider.Route
expectedRoutes []*cloudprovider.Route
expectedNetworkUnavailable []bool
clientset *fake.Clientset
}{
// 2 nodes, routes already there
{
nodes: []api.Node{
node1,
node2,
},
initialRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-02", "node-2", "10.120.1.0/24"},
},
expectedRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-02", "node-2", "10.120.1.0/24"},
},
expectedNetworkUnavailable: []bool{true, true},
clientset: fake.NewSimpleClientset(&api.NodeList{Items: []api.Node{node1, node2}}),
},
// 2 nodes, one route already there
{
nodes: []api.Node{
node1,
node2,
},
initialRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
},
expectedRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-02", "node-2", "10.120.1.0/24"},
},
expectedNetworkUnavailable: []bool{true, true},
clientset: fake.NewSimpleClientset(&api.NodeList{Items: []api.Node{node1, node2}}),
},
// 2 nodes, no routes yet
{
nodes: []api.Node{
node1,
node2,
},
initialRoutes: []*cloudprovider.Route{},
expectedRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-02", "node-2", "10.120.1.0/24"},
},
expectedNetworkUnavailable: []bool{true, true},
clientset: fake.NewSimpleClientset(&api.NodeList{Items: []api.Node{node1, node2}}),
},
// 2 nodes, a few too many routes
{
nodes: []api.Node{
node1,
node2,
},
initialRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-02", "node-2", "10.120.1.0/24"},
{cluster + "-03", "node-3", "10.120.2.0/24"},
{cluster + "-04", "node-4", "10.120.3.0/24"},
},
expectedRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-02", "node-2", "10.120.1.0/24"},
},
expectedNetworkUnavailable: []bool{true, true},
clientset: fake.NewSimpleClientset(&api.NodeList{Items: []api.Node{node1, node2}}),
},
// 2 nodes, 2 routes, but only 1 is right
{
nodes: []api.Node{
node1,
node2,
},
initialRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-03", "node-3", "10.120.2.0/24"},
},
expectedRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
{cluster + "-02", "node-2", "10.120.1.0/24"},
},
expectedNetworkUnavailable: []bool{true, true},
clientset: fake.NewSimpleClientset(&api.NodeList{Items: []api.Node{node1, node2}}),
},
// 2 nodes, one node without CIDR assigned.
{
nodes: []api.Node{
node1,
nodeNoCidr,
},
initialRoutes: []*cloudprovider.Route{},
expectedRoutes: []*cloudprovider.Route{
{cluster + "-01", "node-1", "10.120.0.0/24"},
},
expectedNetworkUnavailable: []bool{true, false},
clientset: fake.NewSimpleClientset(&api.NodeList{Items: []api.Node{node1, nodeNoCidr}}),
},
}
for i, testCase := range testCases {
cloud := &fakecloud.FakeCloud{RouteMap: make(map[string]*fakecloud.FakeRoute)}
for _, route := range testCase.initialRoutes {
fakeRoute := &fakecloud.FakeRoute{}
fakeRoute.ClusterName = cluster
fakeRoute.Route = *route
cloud.RouteMap[route.Name] = fakeRoute
}
routes, ok := cloud.Routes()
if !ok {
t.Error("Error in test: fakecloud doesn't support Routes()")
}
_, cidr, _ := net.ParseCIDR("10.120.0.0/16")
rc := New(routes, testCase.clientset, cluster, cidr)
if err := rc.reconcile(testCase.nodes, testCase.initialRoutes); err != nil {
t.Errorf("%d. Error from rc.reconcile(): %v", i, err)
}
for _, action := range testCase.clientset.Actions() {
if action.GetVerb() == "update" && action.GetResource().Resource == "nodes" {
node := action.(core.UpdateAction).GetObject().(*api.Node)
_, condition := api.GetNodeCondition(&node.Status, api.NodeNetworkUnavailable)
if condition == nil {
t.Errorf("%d. Missing NodeNetworkUnavailable condition for Node %v", i, node.Name)
} else {
check := func(index int) bool {
return (condition.Status == api.ConditionFalse) == testCase.expectedNetworkUnavailable[index]
}
index := -1
for j := range testCase.nodes {
if testCase.nodes[j].Name == node.Name {
index = j
}
}
if index == -1 {
// Something's wrong
continue
}
if !check(index) {
t.Errorf("%d. Invalid NodeNetworkUnavailable condition for Node %v, expected %v, got %v",
i, node.Name, testCase.expectedNetworkUnavailable[index], (condition.Status == api.ConditionFalse))
}
}
}
}
var finalRoutes []*cloudprovider.Route
var err error
timeoutChan := time.After(200 * time.Millisecond)
tick := time.NewTicker(10 * time.Millisecond)
defer tick.Stop()
poll:
for {
select {
case <-tick.C:
if finalRoutes, err = routes.ListRoutes(cluster); err == nil && routeListEqual(finalRoutes, testCase.expectedRoutes) {
break poll
}
case <-timeoutChan:
t.Errorf("%d. rc.reconcile() = %v, routes:\n%v\nexpected: nil, routes:\n%v\n", i, err, flatten(finalRoutes), flatten(testCase.expectedRoutes))
break poll
}
}
}
}
Context("when there is memory pressure", func() {
AfterEach(func() {
// Wait for the memory pressure condition to disappear from the node status before continuing.
By("waiting for the memory pressure condition on the node to disappear before ending the test.")
Eventually(func() error {
nodeList, err := f.Client.Nodes().List(api.ListOptions{})
if err != nil {
return fmt.Errorf("tried to get node list but got error: %v", err)
}
// Assuming that there is only one node, because this is a node e2e test.
if len(nodeList.Items) != 1 {
return fmt.Errorf("expected 1 node, but see %d. List: %v", len(nodeList.Items), nodeList.Items)
}
node := nodeList.Items[0]
_, pressure := api.GetNodeCondition(&node.Status, api.NodeMemoryPressure)
if pressure != nil && pressure.Status == api.ConditionTrue {
return fmt.Errorf("node is still reporting memory pressure condition: %s", pressure)
}
return nil
}, 5*time.Minute, 15*time.Second).Should(BeNil())
// Check available memory after condition disappears, just in case:
// Wait for available memory to decrease to a reasonable level before ending the test.
// This helps prevent interference with tests that start immediately after this one.
By("waiting for available memory to decrease to a reasonable level before ending the test.")
Eventually(func() error {
summary, err := getNodeSummary()
if err != nil {
return err
}
func (rc *RouteController) reconcile(nodes []api.Node, routes []*cloudprovider.Route) error {
// nodeCIDRs maps nodeName->nodeCIDR
nodeCIDRs := make(map[types.NodeName]string)
// routeMap maps routeTargetNode->route
routeMap := make(map[types.NodeName]*cloudprovider.Route)
for _, route := range routes {
routeMap[route.TargetNode] = route
}
wg := sync.WaitGroup{}
rateLimiter := make(chan struct{}, maxConcurrentRouteCreations)
for _, node := range nodes {
// Skip if the node hasn't been assigned a CIDR yet.
if node.Spec.PodCIDR == "" {
continue
}
nodeName := types.NodeName(node.Name)
// Check if we have a route for this node w/ the correct CIDR.
r := routeMap[nodeName]
if r == nil || r.DestinationCIDR != node.Spec.PodCIDR {
// If not, create the route.
route := &cloudprovider.Route{
TargetNode: nodeName,
DestinationCIDR: node.Spec.PodCIDR,
}
nameHint := string(node.UID)
wg.Add(1)
go func(nodeName types.NodeName, nameHint string, route *cloudprovider.Route) {
defer wg.Done()
for i := 0; i < maxRetries; i++ {
startTime := time.Now()
// Ensure that we don't have more than maxConcurrentRouteCreations
// CreateRoute calls in flight.
rateLimiter <- struct{}{}
glog.Infof("Creating route for node %s %s with hint %s, throttled %v", nodeName, route.DestinationCIDR, nameHint, time.Now().Sub(startTime))
err := rc.routes.CreateRoute(rc.clusterName, nameHint, route)
<-rateLimiter
rc.updateNetworkingCondition(nodeName, err == nil)
if err != nil {
glog.Errorf("Could not create route %s %s for node %s after %v: %v", nameHint, route.DestinationCIDR, nodeName, time.Now().Sub(startTime), err)
} else {
glog.Infof("Created route for node %s %s with hint %s after %v", nodeName, route.DestinationCIDR, nameHint, time.Now().Sub(startTime))
return
}
}
}(nodeName, nameHint, route)
} else {
// Update condition only if it doesn't reflect the current state.
_, condition := api.GetNodeCondition(&node.Status, api.NodeNetworkUnavailable)
if condition == nil || condition.Status != api.ConditionFalse {
rc.updateNetworkingCondition(types.NodeName(node.Name), true)
}
}
nodeCIDRs[nodeName] = node.Spec.PodCIDR
}
for _, route := range routes {
if rc.isResponsibleForRoute(route) {
// Check if this route applies to a node we know about & has correct CIDR.
if nodeCIDRs[route.TargetNode] != route.DestinationCIDR {
wg.Add(1)
// Delete the route.
go func(route *cloudprovider.Route, startTime time.Time) {
glog.Infof("Deleting route %s %s", route.Name, route.DestinationCIDR)
if err := rc.routes.DeleteRoute(rc.clusterName, route); err != nil {
glog.Errorf("Could not delete route %s %s after %v: %v", route.Name, route.DestinationCIDR, time.Now().Sub(startTime), err)
} else {
glog.Infof("Deleted route %s %s after %v", route.Name, route.DestinationCIDR, time.Now().Sub(startTime))
}
wg.Done()
}(route, time.Now())
}
}
}
wg.Wait()
return nil
}