Merge pull request #337 from nais/haproxy_http2_fix

fix(hookd): recover clusters after dispatch stream failure

Author: Muni10

.envrc

@@ -0,0 +1 @@
+use flake

.gitignore

@@ -21,3 +21,4 @@
 .env
 
 /.testbin/
+/.direnv/

flake.lock

@@ -0,0 +1,75 @@
+{
+  inputs = {
+    flake-utils.url = "github:numtide/flake-utils";
+    nixpkgs.url = "nixpkgs/nixos-unstable";
+    treefmt-nix = {
+      url = "github:numtide/treefmt-nix";
+      inputs.nixpkgs.follows = "nixpkgs";
+    };
+  };
+
+  outputs =
+    inputs:
+    inputs.flake-utils.lib.eachDefaultSystem (
+      system:
+      let
+        pkgs = import inputs.nixpkgs {
+          localSystem = {
+            inherit system;
+          };
+        };
+        treefmt = inputs.treefmt-nix.lib.evalModule pkgs {
+          projectRootFile = "flake.nix";
+          programs.nixfmt.enable = true;
+        };
+        envtest-bins = pkgs.symlinkJoin {
+          name = "envtest-bins";
+          paths = [
+            pkgs.etcd
+            pkgs.kubernetes
+          ];
+        };
+        deploy = pkgs.buildGoModule {
+          pname = "deploy";
+          version = "0.0.0";
+          src = ./.;
+          vendorHash = "sha256-afFOb7DpB4gGyFErwM3lROMU2E1GVlT7+nSLU4zAV8E=";
+
+          subPackages = [
+            "cmd/crypt"
+            "cmd/deploy"
+            "cmd/deployd"
+            "cmd/hookd"
+            "cmd/leakdetect"
+          ];
+
+          nativeBuildInputs = with pkgs; [
+            protobuf
+            protoc-gen-go
+            protoc-gen-go-grpc
+          ];
+
+          doCheck = false;
+        };
+      in
+      {
+        packages.default = deploy;
+
+        devShells.default = pkgs.mkShell {
+          inputsFrom = [
+            deploy
+            envtest-bins
+          ];
+          shellHook = ''
+            export KUBEBUILDER_ASSETS="${envtest-bins}/bin"
+            mkdir -p .testbin
+            ln -sfn ${pkgs.etcd}/bin/etcd .testbin/etcd
+            ln -sfn ${pkgs.kubernetes}/bin/kube-apiserver .testbin/kube-apiserver
+            ln -sfn ${pkgs.kubernetes}/bin/kubectl .testbin/kubectl
+          '';
+        };
+
+        formatter = treefmt.config.build.wrapper;
+      }
+    );
+}

flake.nix

@@ -129,6 +129,9 @@ func (s *dispatchServer) Deployments(opts *pb.GetDeploymentOpts, stream pb.Dispa
 		case req := <-c:
 			err := stream.Send(req.request)
 			req.wait <- err
+			if err != nil {
+				return err
+			}
 		case <-time.After(30 * time.Minute):
 			log.Warnf("Connection from cluster '%s' timed out", opts.Cluster)
 			return fmt.Errorf("timeout")

pkg/grpc/dispatchserver/dispatchserver.go

@@ -16,10 +16,84 @@ import (
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/credentials/insecure"
+	"google.golang.org/grpc/metadata"
 	"google.golang.org/grpc/status"
 	"google.golang.org/grpc/test/bufconn"
 )
 
+type erringDeploymentsStream struct {
+	ctx context.Context
+}
+
+func (s *erringDeploymentsStream) Send(*pb.DeploymentRequest) error {
+	return status.Error(codes.Internal, "stream terminated by RST_STREAM with error code: INTERNAL_ERROR")
+}
+
+func (s *erringDeploymentsStream) Context() context.Context { return s.ctx }
+
+func (s *erringDeploymentsStream) SetHeader(metadata.MD) error { return nil }
+
+func (s *erringDeploymentsStream) SendHeader(metadata.MD) error { return nil }
+
+func (s *erringDeploymentsStream) SetTrailer(metadata.MD) {}
+
+func (s *erringDeploymentsStream) SendMsg(any) error { return nil }
+
+func (s *erringDeploymentsStream) RecvMsg(any) error { return nil }
+
+func TestDeploymentsUnregistersClusterWhenSendFails(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	_, _ = telemetry.New(ctx, "test", "")
+
+	deploymentStore := database.MockDeploymentStore{}
+	deploymentStore.On("HistoricDeployments", mock.Anything, mock.Anything, mock.Anything).Return(nil, nil)
+
+	mockApiClients, _ := apiclient.NewMockClient(t)
+	ds := New(&deploymentStore, mockApiClients.Deployments()).(*dispatchServer)
+
+	stream := &erringDeploymentsStream{ctx: ctx}
+	done := make(chan error, 1)
+	go func() {
+		done <- ds.Deployments(&pb.GetDeploymentOpts{Cluster: "dev", StartupTime: pb.TimeAsTimestamp(time.Now())}, stream)
+	}()
+
+	requireEventually(t, time.Second, func() bool {
+		return len(ds.onlineClusters()) == 1
+	})
+
+	err := ds.SendDeploymentRequest(ctx, &pb.DeploymentRequest{Cluster: "dev"})
+	if status.Code(err) != codes.Internal {
+		t.Fatalf("expected send error to be propagated as Internal, got %v", err)
+	}
+
+	requireEventually(t, time.Second, func() bool {
+		return len(ds.onlineClusters()) == 0
+	})
+
+	select {
+	case err := <-done:
+		if err == nil {
+			t.Fatal("expected Deployments to exit with send error")
+		}
+	case <-time.After(time.Second):
+		t.Fatal("Deployments did not exit after stream Send failed")
+	}
+}
+
+func requireEventually(t *testing.T, timeout time.Duration, condition func() bool) {
+	t.Helper()
+
+	deadline := time.Now().Add(timeout)
+	for time.Now().Before(deadline) {
+		if condition() {
+			return
+		}
+		time.Sleep(10 * time.Millisecond)
+	}
+	t.Fatal("condition was not satisfied before timeout")
+}
+
 func bufDialer(b *bufconn.Listener) func(context.Context, string) (net.Conn, error) {
 	return func(context.Context, string) (net.Conn, error) {
 		return b.Dial()