Devstack cancellations (#3487)

When exiting devstack, there are often errors logged that are
unnecessary, or in some cases delay when the process actuall exits.

This commit checks some obvious context cancellation errors and gives
the code a chance to exit without logging and continuing.

Should also fix #3484

pkg/devstack/devstack.go

@@ -136,6 +136,7 @@ func Setup(
 			if err != nil {
 				return nil, fmt.Errorf("failed to get ipfs swarm addresses: %w", err)
 			}
+
 			// Only use a single address as libp2p seems to have concurrency issues, like two nodes not able to finish
 			// connecting/joining topics, when using multiple addresses for a single host.
 			// All the IPFS nodes are running within the same process, so connecting over localhost will be fine.
@@ -373,12 +374,6 @@ func (stack *DevStack) PrintNodeInfo(ctx context.Context, fsRepo *repo.FsRepo, c
 	devStackIPFSSwarmAddress := ""
 	var devstackPeerAddrs []string
 
-	// TODO remove this it's wrong and never printed, nothing sets the env vars its printing
-	logString += `
------------------------------------------
------------------------------------------
-`
-
 	requesterOnlyNodes := 0
 	computeOnlyNodes := 0
 	hybridNodes := 0

pkg/ipfs/client.go

@@ -128,6 +128,10 @@ func (cl Client) SwarmAddresses(ctx context.Context) ([]string, error) {
 		return nil, err
 	}
 
+	if len(multiAddresses) == 0 {
+		return nil, fmt.Errorf("no swarm addresses found")
+	}
+
 	// It's common for callers to this function to use the result to connect to another IPFS node.
 	// This sorts the addresses so IPv4 localhost is first, with the aim of using the localhost connection during tests
 	// and so avoid any unneeded network hops. Other callers to this either sort the list themselves or just output the

pkg/ipfs/node.go

@@ -83,6 +83,10 @@ func NewNodeWithConfig(ctx context.Context, cm *system.CleanupManager, cfg types
 
 	// TODO if cfg.PrivateInternal is true do we actually want to connect to peers?
 	if err = connectToPeers(ctx, api, ipfsNode, cfg.GetSwarmAddresses()); err != nil {
+		if err == context.Canceled {
+			log.Ctx(ctx).Debug().Msg("gracefully shutting down ipfs node due to context cancellation")
+			return nil, nil
+		}
 		log.Ctx(ctx).Error().Msgf("ipfs node failed to connect to peers: %s", err)
 	}
 
@@ -343,6 +347,9 @@ func connectToPeers(ctx context.Context, api icore.CoreAPI, node *core.IpfsNode,
 		go func(peerInfo peer.AddrInfo) {
 			defer wg.Done()
 			if err := api.Swarm().Connect(ctx, peerInfo); err != nil {
+				if err == context.Canceled {
+					return // context was canceled, nothing to do
+				}
 				anyErr = err
 				log.Ctx(ctx).Debug().Err(err).Msgf("failed to connect to ipfs peer %s, skipping", peerInfo.ID)
 			}

pkg/routing/node_info_publisher.go

@@ -96,22 +96,26 @@ func (n *NodeInfoPublisher) publishBackgroundTask(ctx context.Context, interval
 	ticker := time.NewTicker(interval)
 	for {
 		select {
+		case <-ctx.Done():
+			return
 		case <-ticker.C:
 			func() {
 				ctx, span := system.NewSpan(ctx, system.GetTracer(), "pkg/routing.NodeInfoPublisher.publishBackgroundTask") //nolint:govet
 				defer span.End()
 
 				err := n.Publish(ctx)
 				if err != nil {
-					log.Ctx(ctx).Err(err).Msg("failed to publish node info")
+					if err == context.Canceled {
+						log.Ctx(ctx).Debug().Msg("gracefully shutting down ipfs node due to context cancellation")
+					} else {
+						log.Ctx(ctx).Err(err).Msg("failed to publish node info")
+					}
 				}
 			}()
 		case <-n.stopChannel:
 			log.Ctx(ctx).Debug().Msg("stopped publishing node info")
 			ticker.Stop()
 			return
-		case <-ctx.Done():
-			return
 		}
 	}
 }