Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func CreateClusterLoadAssignment ¶
func CreateClusterLoadAssignment(edsServiceName string, servingPort uint32, nodeHash string, localityPriorityMapper LocalityPriorityMapper, endpoints []applications.ApplicationEndpoints) *endpointv3.ClusterLoadAssignment
CreateClusterLoadAssignment for EDS. `edsServiceName` must match the `ServiceName` in the `EDSClusterConfig` in the CDS Cluster resource. [gRFC A27]: https://github.com/grpc/proposal/blob/972b69ab1f0f7f6079af81a8c2b8a01a15ce3bec/A27-xds-global-load-balancing.md#clusterloadassignment-proto
Types ¶
type FixedLocalityPriority ¶
type FixedLocalityPriority struct{}
FixedLocalityPriority returns an empty map. Lookups in the map will always return 0 as the value, so all localities can be assigned the highest priority.
func (FixedLocalityPriority) BuildPriorityMap ¶
func (f FixedLocalityPriority) BuildPriorityMap(_ string, _ []string) map[string]uint32
type LocalityMatch ¶
type LocalityMatch int
LocalityMatch defines the priority order of matching or part-matching locality. In other words, the priority order is as follows:
- same zone
- same region
- same super-region
- same multi-region
- other
const ( Zone LocalityMatch = iota Region SuperRegion MultiRegion Other )
type LocalityPriorityByZone ¶
type LocalityPriorityByZone struct{}
LocalityPriorityByZone determines EDS ClusterLoadAssignment locality priorites, based on the zone of the requesting node.
Priorities start from 0 (highest), and must increase without any gaps. Multiple localities can share the same priority.
Zone and region are names of cloud provider locations, e.g., the `us-central1-f` zone, and the `us-central1` region.
Super-region is an invented term for this implementation. It is essentially the cloud provider region, excluding the numeric suffix. E.g., for the `us-west1` region, the super-region is `us-west`. This means that the `us-west1` and `us-west2` regions are in the same super-region. The intention is to provide a next level of priority, after considering zone and region. E.g., if traffic originates from `us-west1`, but no healthy endpoints are available in that region, we may prefer to send traffic to a nearby region, such as, e.g., `us-west2`, before considering other regions that are geographically more distant, e.g., `us-east1`.
Multi-region is an invented term for this implementation. It is essentially the first part of the cloud provider region name, up to the first hyphen (`-`). E.g., for the region `us-east1`, the multi-region is `us`. This means that the `us-west1` and `us-east1` regions are in the same multi-region. The intention is to provide a next level of priority, after considering zone, region, and super-region. E.g., if traffic originates from `us-west1`, but no healthy endpoints are available in any of the `us-west1*` regions, we may prefer to send traffic to another region on the same continent, such as, e.g., `us-east1`, before considering regions on other continents, e.g., `europe-west1`.
func (LocalityPriorityByZone) BuildPriorityMap ¶
func (l LocalityPriorityByZone) BuildPriorityMap(nodeZone string, zonesToPrioritize []string) map[string]uint32
BuildPriorityMap constructs the priority map for the provided zones, based on the zone of the requesting node. Assumption: The nodeHash value (the first argument) is the zone name of the requesting node.
type LocalityPriorityMapper ¶
type LocalityPriorityMapper interface {
// BuildPriorityMap constructs the priority map for the provided zones, based on the node hash.
BuildPriorityMap(nodeHash string, zones []string) map[string]uint32
}
LocalityPriorityMapper determines EDS ClusterLoadAssignment locality priorites.
Source Files