You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Context
Currently, admiral generated SEs point to K8s service names which causes:
i) Istio subset routing to not work
ii) The load balancing doesn't work which impacts protocols like gRPC since the source pods do not directly connect to the destination pods but the clusterIps of the servic as admiral generated SE points to k8s service endpoint.
Proposed solution
Istio 1.6 and higher provides:
i) A new CR called WorkloadEntry
ii) WorkloadSelector on SE to reference the WorkloadEntry
Using these constructs, the SE generated with the local k8s service name can be replaced with a WorkloadEntry and a WorkloadSelector with appropriate labels to deliver the right EDS.
The text was updated successfully, but these errors were encountered:
We found that using headless services (clusterIp: None) will solve the problem of load balancing across pods. We will be deferring the support for workload entry until it graduates with all the bug fixes resolved. Refer to this comment for more information.
Context
Currently, admiral generated SEs point to K8s service names which causes:
i) Istio subset routing to not work
ii) The load balancing doesn't work which impacts protocols like gRPC since the source pods do not directly connect to the destination pods but the clusterIps of the servic as admiral generated SE points to k8s service endpoint.
Proposed solution
Istio 1.6 and higher provides:
i) A new CR called WorkloadEntry
ii) WorkloadSelector on SE to reference the WorkloadEntry
Using these constructs, the SE generated with the local k8s service name can be replaced with a WorkloadEntry and a WorkloadSelector with appropriate labels to deliver the right EDS.
The text was updated successfully, but these errors were encountered: