Reach stellar-core HTTP via pod-exec when --core-http-via-pod-exec is set

src/App/Program.fs

@@ -46,6 +46,7 @@ type MissionOptions
         ingressInternalDomain: string,
         ingressExternalHost: string option,
         ingressExternalPort: int,
+        metricsViaClusterDns: bool,
         exportToPrometheus: bool,
         probeTimeout: int,
         missions: string seq,
@@ -178,6 +179,12 @@ type MissionOptions
              Default = 80)>]
     member self.IngressExternalPort = ingressExternalPort
 
+    [<Option("metrics-via-cluster-dns",
+             HelpText = "Scrape post-mission metrics directly from the per-pod svc.cluster.local DNS name instead of through the ingress hostname. Use when the ingress hostname (e.g. <nonce>.local) does not resolve from where SSC runs but cluster DNS does.",
+             Required = false,
+             Default = false)>]
+    member self.MetricsViaClusterDns = metricsViaClusterDns
+
     [<Option("export-to-prometheus", HelpText = "Whether to export core metrics to prometheus")>]
     member self.ExportToPrometheus : bool = exportToPrometheus
 
@@ -779,6 +786,7 @@ let main argv =
                                ingressInternalDomain = mission.IngressInternalDomain
                                ingressExternalHost = mission.IngressExternalHost
                                ingressExternalPort = mission.IngressExternalPort
+                               metricsViaClusterDns = mission.MetricsViaClusterDns
                                exportToPrometheus = mission.ExportToPrometheus
                                probeTimeout = mission.ProbeTimeout
                                coreResources = SmallTestResources

src/FSLibrary.Tests/Tests.fs

@@ -58,6 +58,7 @@ let ctx : MissionContext =
       ingressInternalDomain = "local"
       ingressExternalHost = None
       ingressExternalPort = 80
+      metricsViaClusterDns = false
       exportToPrometheus = false
       probeTimeout = 10
       coreResources = SmallTestResources

src/FSLibrary/StellarCoreHTTP.fs

@@ -392,6 +392,17 @@ type Peer with
         let url = self.URL path
         Http.RequestString(url, headers = self.Headers)
 
+    // URL that reaches stellar-core's admin HTTP endpoint directly via the
+    // per-pod svc.cluster.local DNS name, bypassing the ingress. The ingress
+    // hostname (<nonce>.<ingress-internal-domain>, e.g. <nonce>.local) does not
+    // resolve in all cluster-DNS environments (for example non-SDF k3s
+    // clusters), whereas the per-pod service DNS name resolves wherever cluster
+    // DNS is reachable. See https://github.com/stellar/supercluster/issues/399.
+    member self.ClusterDnsURL(path: string) : string =
+        sprintf "http://%s:%d/%s" self.DnsName.StringName StellarCoreCfg.CfgVal.httpPort path
+
+    member self.fetchFromClusterDns(path: string) : string = Http.RequestString(self.ClusterDnsURL path)
+
     member self.GetState() = self.GetInfo().State
 
     member self.GetStatusOrState() : string =
@@ -401,7 +412,16 @@ type Peer with
     member self.GetMetrics() : Metrics.Metrics =
         WebExceptionRetry DefaultRetry (fun _ -> Metrics.Parse(self.fetch "metrics").Metrics)
 
-    member self.GetRawMetrics() = WebExceptionRetry DefaultRetry (fun _ -> self.fetch "metrics")
+    // Post-mission metrics are scraped through the ingress by default. When
+    // --metrics-via-cluster-dns is set they are instead scraped directly from
+    // the per-pod svc.cluster.local DNS name (see ClusterDnsURL), so the
+    // teardown dump succeeds in environments where the ingress hostname (e.g.
+    // <nonce>.local) does not resolve from where SSC runs but cluster DNS does.
+    member self.GetRawMetrics() =
+        if self.networkCfg.missionContext.metricsViaClusterDns then
+            WebExceptionRetry DefaultRetry (fun _ -> self.fetchFromClusterDns "metrics")
+        else
+            WebExceptionRetry DefaultRetry (fun _ -> self.fetch "metrics")
 
     member self.GetInfo() : Info.Info =
         WebExceptionRetry

src/FSLibrary/StellarMissionContext.fs

@@ -59,6 +59,7 @@ type MissionContext =
       ingressInternalDomain: string
       ingressExternalHost: string option
       ingressExternalPort: int
+      metricsViaClusterDns: bool
       exportToPrometheus: bool
       probeTimeout: int
       coreResources: CoreResources