about summary refs log tree commit diff
diff options
context:
space:
mode:
authorFlorian Klink <flokli@flokli.de>2023-09-18T13·51+0300
committerclbot <clbot@tvl.fyi>2023-09-18T14·34+0000
commit6c586bc2a7a49755b4b2566c034bc1affc92011d (patch)
tree826be0c654fa0430b8abed47cb154a2d0bfc4749
parent02aed32bf2627a969c15f358737abf5acb697dd4 (diff)
feat(tvix/nar-bridge): graceful shutdown r/6616
This gives existing clients 30s to finish their requests after receiving
an interrupt.

Change-Id: Ia9b0e662fd1ffbbb6c2d03f3dd6548b13cf3d241
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9365
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
-rw-r--r--tvix/nar-bridge/cmd/nar_bridge/main.go34
-rw-r--r--tvix/nar-bridge/pkg/server/server.go12
2 files changed, 30 insertions, 16 deletions
diff --git a/tvix/nar-bridge/cmd/nar_bridge/main.go b/tvix/nar-bridge/cmd/nar_bridge/main.go
index 482012f39b5f..a1732433c050 100644
--- a/tvix/nar-bridge/cmd/nar_bridge/main.go
+++ b/tvix/nar-bridge/cmd/nar_bridge/main.go
@@ -1,8 +1,10 @@
 package main
 
 import (
+	"context"
 	"os"
 	"os/signal"
+	"time"
 
 	"github.com/alecthomas/kong"
 
@@ -33,25 +35,17 @@ func main() {
 	}
 	logrus.SetLevel(logLevel)
 
-	c := make(chan os.Signal, 1)
-	signal.Notify(c, os.Interrupt)
-
-	go func() {
-		for range c {
-			log.Info("Received Signal, shutting down…")
-			os.Exit(1)
-		}
-	}()
+	ctx, stop := signal.NotifyContext(context.Background(), os.Interrupt)
+	defer stop()
 
 	// connect to tvix-store
 	log.Debugf("Dialing to %v", cli.StoreAddr)
-	conn, err := grpc.Dial(cli.StoreAddr, grpc.WithTransportCredentials(insecure.NewCredentials()))
+	conn, err := grpc.DialContext(ctx, cli.StoreAddr, grpc.WithTransportCredentials(insecure.NewCredentials()))
 	if err != nil {
 		log.Fatalf("did not connect: %v", err)
 	}
 	defer conn.Close()
 
-	log.Printf("Starting nar-bridge at %v", cli.ListenAddr)
 	s := server.New(
 		storev1pb.NewDirectoryServiceClient(conn),
 		storev1pb.NewBlobServiceClient(conn),
@@ -60,9 +54,21 @@ func main() {
 		30,
 	)
 
-	err = s.ListenAndServe(cli.ListenAddr)
-	if err != nil {
-		log.Error("Server failed: %w", err)
+	log.Printf("Starting nar-bridge at %v", cli.ListenAddr)
+	go s.ListenAndServe(cli.ListenAddr)
+
+	// listen for the interrupt signal.
+	<-ctx.Done()
+
+	// Restore default behaviour on the interrupt signal
+	stop()
+	log.Info("Received Signal, shutting down, press Ctl+C again to force.")
+
+	timeoutCtx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+
+	if s.Shutdown(timeoutCtx); err != nil {
+		log.WithError(err).Warn("failed to shutdown")
 		os.Exit(1)
 	}
 }
diff --git a/tvix/nar-bridge/pkg/server/server.go b/tvix/nar-bridge/pkg/server/server.go
index 0de38c84dfee..f58842bfa792 100644
--- a/tvix/nar-bridge/pkg/server/server.go
+++ b/tvix/nar-bridge/pkg/server/server.go
@@ -1,6 +1,7 @@
 package server
 
 import (
+	"context"
 	"fmt"
 	"net/http"
 	"sync"
@@ -13,6 +14,7 @@ import (
 )
 
 type Server struct {
+	srv     *http.Server
 	handler chi.Router
 
 	directoryServiceClient storev1pb.DirectoryServiceClient
@@ -73,8 +75,14 @@ func New(
 	return s
 }
 
+func (s *Server) Shutdown(ctx context.Context) error {
+	return s.srv.Shutdown(ctx)
+}
+
+// ListenAndServer starts the webserver, and waits for it being closed or
+// shutdown, after which it'll return ErrServerClosed.
 func (s *Server) ListenAndServe(addr string) error {
-	srv := &http.Server{
+	s.srv = &http.Server{
 		Addr:         addr,
 		Handler:      s.handler,
 		ReadTimeout:  500 * time.Second,
@@ -82,5 +90,5 @@ func (s *Server) ListenAndServe(addr string) error {
 		IdleTimeout:  500 * time.Second,
 	}
 
-	return srv.ListenAndServe()
+	return s.srv.ListenAndServe()
 }