util

Documentation

Index

• func HandleMultiClusterRequest(sourceWlNs, sourceWl, destSvcNs, destSvcName string) (string, string)
• func HandleResponseCode(protocol, httpResponseCode string, grpcResponseStatusOk bool, ...) string
• func IsBadDestTelemetry(svc, svcName, wl string) bool
• func IsBadSourceTelemetry(ns, wl, app string) bool

Functions

func HandleMultiClusterRequest

func HandleMultiClusterRequest(sourceWlNs, sourceWl, destSvcNs, destSvcName string) (string, string)

HandleMultiClusterRequest ensures the proper destination service workload and name for requests forwarded from another cluster (via a ServiceEntry).

Given a request from clusterA to clusterB, clusterA will generate source telemetry (from the source workload to the service entry) and clusterB will generate destination telemetry (from unknown to the destination workload). If this is the destination telemetry the destination_service_name label will be set to the service entry host, which is required to have the form <name>.<namespace>.global where name and namespace correspond to the remote service’s name and namespace respectively. In this situation we alter the request in two ways:

First, we reset destSvcName to <name> in order to unify remote and local requests to the service. By doing this the graph will show only one <service> node instead of having a node for both <service> and <name>.<namespace>.global which in practice, are the same.

Second, we reset destSvcNs to <namespace>. We want destSvcNs to be set to the namespace of the remote service's namespace. But in practice it will be set to the namespace (on clusterA) where the servieEntry is defined. This is not useful for the visualization, and so we replace it here. Note that <namespace> should be equivalent to the value set for destination_workload_namespace, we just use <namespace> for convenience, we have it here.

All of this is only done if source workload is "unknown", which is what indicates that this represents the destination telemetry on clusterB, and if the destSvcName is in the MC format. When the source workload IS known the traffic it should be representing the clusterA traffic to the ServiceEntry and being routed out of the cluster. That use case is handled in the service_entry.go file.

func HandleResponseCode

func HandleResponseCode(protocol, httpResponseCode string, grpcResponseStatusOk bool, grpcResponseStatus string) string

HandleResponseCode returns either the HTTP response code or the GRPC response status. GRPC response status was added upstream in Istio 1.5 and downstream OSSM 1.1. We support it here in a backward compatible way. When protocol is not GRPC, or if the version running does not supply the GRPC status, just return the HTTP code. Also return the HTTP code In the rare case that protocol is GRPC but the HTTP transport fails. (I have never seen this happen). Otherwise, return the GRPC status.

func IsBadDestTelemetry

func IsBadDestTelemetry(svc, svcName, wl string) bool

IsBadDestTelemetry tests for known issues in generated telemetry given indicative label values.

1. During pod lifecycle changes incomplete telemetry may be generated that results in destSvc == destSvcName and no dest workload, where destSvc[Name] is in the form of an IP address.
2. no more conditions known

func IsBadSourceTelemetry

func IsBadSourceTelemetry(ns, wl, app string) bool

IsBadSourceTelemetry tests for known issues in generated telemetry given indicative label values. 1) source namespace is provided with neither workload nor app 2) no more conditions known