github/kubernetes
4
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Ubernetes Lite: Volumes can dictate zone scheduling

Browse Source 
For AWS EBS, a volume can only be attached to a node in the same AZ.
The scheduler must therefore detect if a volume is being attached to a
pod, and ensure that the pod is scheduled on a node in the same AZ as
the volume.

So that the scheduler need not query the cloud provider every time, and
to support decoupled operation (e.g. bare metal) we tag the volume with
our placement labels.  This is done automatically by means of an
admission controller on AWS when a PersistentVolume is created backed by
an EBS volume.

Support for tagging GCE PVs will follow.

Pods that specify a volume directly (i.e. without using a
PersistentVolumeClaim) will not currently be scheduled correctly (i.e.
they will be scheduled without zone-awareness).
This commit is contained in:
Justin Santa Barbara
2015-11-29 14:00:49 -05:00
parent 7743a4ca89
commit f9a6ac077e
15 changed files with 536 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
plugin/pkg/scheduler/factory/factory.go
View File

@@ -58,6 +58,10 @@ type ConfigFactory struct {
PodLister algorithm.PodLister
// a means to list all nodes
NodeLister *cache.StoreToNodeLister
// a means to list all PersistentVolumes
PVLister *cache.StoreToPVFetcher
// a means to list all PersistentVolumeClaims
PVCLister *cache.StoreToPVCFetcher
// a means to list all services
ServiceLister *cache.StoreToServiceLister
// a means to list all controllers
@@ -85,6 +89,8 @@ func NewConfigFactory(client *client.Client, rateLimiter util.RateLimiter, sched
ScheduledPodLister: &cache.StoreToPodLister{},
// Only nodes in the "Ready" condition with status == "True" are schedulable
NodeLister: &cache.StoreToNodeLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
PVLister: &cache.StoreToPVFetcher{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
PVCLister: &cache.StoreToPVCFetcher{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ServiceLister: &cache.StoreToServiceLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
ControllerLister: &cache.StoreToReplicationControllerLister{Store: cache.NewStore(cache.MetaNamespaceKeyFunc)},
StopEverything: make(chan struct{}),
@@ -188,6 +194,8 @@ func (f *ConfigFactory) CreateFromKeys(predicateKeys, priorityKeys sets.String,
// All fit predicates only need to consider schedulable nodes.
NodeLister: f.NodeLister.NodeCondition(getNodeConditionPredicate()),
NodeInfo: &predicates.CachedNodeInfo{f.NodeLister},
PVInfo: f.PVLister,
PVCInfo: f.PVCLister,
}
predicateFuncs, err := getFitPredicateFunctions(predicateKeys, pluginArgs)
if err != nil {
@@ -209,6 +217,11 @@ func (f *ConfigFactory) CreateFromKeys(predicateKeys, priorityKeys sets.String,
// Nodes may be listed frequently, so provide a local up-to-date cache.
cache.NewReflector(f.createNodeLW(), &api.Node{}, f.NodeLister.Store, 0).RunUntil(f.StopEverything)
// Watch PVs & PVCs
// They may be listed frequently for scheduling constraints, so provide a local up-to-date cache.
cache.NewReflector(f.createPersistentVolumeLW(), &api.PersistentVolume{}, f.PVLister.Store, 0).RunUntil(f.StopEverything)
cache.NewReflector(f.createPersistentVolumeClaimLW(), &api.PersistentVolumeClaim{}, f.PVCLister.Store, 0).RunUntil(f.StopEverything)
// Watch and cache all service objects. Scheduler needs to find all pods
// created by the same services or ReplicationControllers, so that it can spread them correctly.
// Cache this locally.
@@ -303,6 +316,16 @@ func (factory *ConfigFactory) createNodeLW() *cache.ListWatch {
return cache.NewListWatchFromClient(factory.Client, "nodes", api.NamespaceAll, fields)
}
// createPersistentVolumeLW returns a cache.ListWatch that gets all changes to persistentVolumes.
func (factory *ConfigFactory) createPersistentVolumeLW() *cache.ListWatch {
return cache.NewListWatchFromClient(factory.Client, "persistentVolumes", api.NamespaceAll, fields.ParseSelectorOrDie(""))
}
// createPersistentVolumeClaimLW returns a cache.ListWatch that gets all changes to persistentVolumeClaims.
func (factory *ConfigFactory) createPersistentVolumeClaimLW() *cache.ListWatch {
return cache.NewListWatchFromClient(factory.Client, "persistentVolumeClaims", api.NamespaceAll, fields.ParseSelectorOrDie(""))
}
// Returns a cache.ListWatch that gets all changes to services.
func (factory *ConfigFactory) createServiceLW() *cache.ListWatch {
return cache.NewListWatchFromClient(factory.Client, "services", api.NamespaceAll, fields.ParseSelectorOrDie(""))