bring back minor DNS cache for k8s setups (#19341)

k8s as it stands is flaky in DNS lookups,
bring this change back such that we can
cache DNS atleast for 30secs TTL.

cmd/common-main.go

@@ -66,9 +66,11 @@ import (
 )
 
 // serverDebugLog will enable debug printing
-var serverDebugLog = env.Get("_MINIO_SERVER_DEBUG", config.EnableOff) == config.EnableOn
-
-var currentReleaseTime time.Time
+var (
+	serverDebugLog     = env.Get("_MINIO_SERVER_DEBUG", config.EnableOff) == config.EnableOn
+	currentReleaseTime time.Time
+	orchestrated       = IsKubernetes() || IsDocker()
+)
 
 func init() {
 	if runtime.GOOS == "windows" {
@@ -494,7 +496,11 @@ func runDNSCache(ctx *cli.Context) {
 	dnsTTL := ctx.Duration("dns-cache-ttl")
 	// Check if we have configured a custom DNS cache TTL.
 	if dnsTTL <= 0 {
-		dnsTTL = 10 * time.Minute
+		if orchestrated {
+			dnsTTL = 30 * time.Second
+		} else {
+			dnsTTL = 10 * time.Minute
+		}
 	}
 
 	// Call to refresh will refresh names in cache.
@@ -757,12 +763,7 @@ func serverHandleEnvVars() {
 		for _, endpoint := range minioEndpoints {
 			if net.ParseIP(endpoint) == nil {
 				// Checking if the IP is a DNS entry.
-				lookupHost := globalDNSCache.LookupHost
-				if IsKubernetes() || IsDocker() {
-					lookupHost = net.DefaultResolver.LookupHost
-				}
-
-				addrs, err := lookupHost(GlobalContext, endpoint)
+				addrs, err := globalDNSCache.LookupHost(GlobalContext, endpoint)
 				if err != nil {
 					logger.FatalIf(err, "Unable to initialize MinIO server with [%s] invalid entry found in MINIO_PUBLIC_IPS", endpoint)
 				}

cmd/endpoint.go

@@ -590,8 +590,6 @@ func hostResolveToLocalhost(endpoint Endpoint) bool {
 
 // UpdateIsLocal - resolves the host and discovers the local host.
 func (endpoints Endpoints) UpdateIsLocal() error {
-	orchestrated := IsDocker() || IsKubernetes()
-
 	var epsResolved int
 	var foundLocal bool
 	resolvedList := make([]bool, len(endpoints))
@@ -775,8 +773,6 @@ type PoolEndpointList []Endpoints
 
 // UpdateIsLocal - resolves all hosts and discovers which are local
 func (p PoolEndpointList) UpdateIsLocal() error {
-	orchestrated := IsDocker() || IsKubernetes()
-
 	var epsResolved int
 	var epCount int
 
@@ -1034,7 +1030,6 @@ func CreatePoolEndpoints(serverAddr string, poolsLayout ...poolDisksLayout) ([]E
 			}
 		}
 
-		orchestrated := IsKubernetes() || IsDocker()
 		reverseProxy := (env.Get("_MINIO_REVERSE_PROXY", "") != "") && ((env.Get("MINIO_CI_CD", "") != "") || (env.Get("CI", "") != ""))
 		// If not orchestrated
 		// and not setup in reverse proxy

cmd/grid.go

@@ -35,13 +35,9 @@ var globalGrid atomic.Pointer[grid.Manager]
 var globalGridStart = make(chan struct{})
 
 func initGlobalGrid(ctx context.Context, eps EndpointServerPools) error {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = nil
-	}
 	hosts, local := eps.GridHosts()
 	g, err := grid.NewManager(ctx, grid.ManagerOptions{
-		Dialer:       grid.ContextDialer(xhttp.DialContextWithLookupHost(lookupHost, xhttp.NewInternodeDialContext(rest.DefaultTimeout, globalTCPOptions))),
+		Dialer:       grid.ContextDialer(xhttp.DialContextWithLookupHost(globalDNSCache.LookupHost, xhttp.NewInternodeDialContext(rest.DefaultTimeout, globalTCPOptions))),
 		Local:        local,
 		Hosts:        hosts,
 		AddAuth:      newCachedAuthToken(),

cmd/net.go

@@ -98,12 +98,7 @@ func mustGetLocalIP6() (ipList set.StringSet) {
 
 // getHostIP returns IP address of given host.
 func getHostIP(host string) (ipList set.StringSet, err error) {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = net.DefaultResolver.LookupHost
-	}
-
-	addrs, err := lookupHost(GlobalContext, host)
+	addrs, err := globalDNSCache.LookupHost(GlobalContext, host)
 	if err != nil {
 		return ipList, err
 	}

cmd/server-main.go

@@ -141,9 +141,14 @@ var ServerFlags = []cli.Flag{
 	},
 	cli.DurationFlag{
 		Name:   "dns-cache-ttl",
-		Usage:  "custom DNS cache TTL for baremetal setups",
+		Usage:  "custom DNS cache TTL",
 		Hidden: true,
-		Value:  10 * time.Minute,
+		Value: func() time.Duration {
+			if orchestrated {
+				return 30 * time.Second
+			}
+			return 10 * time.Minute
+		}(),
 		EnvVar: "MINIO_DNS_CACHE_TTL",
 	},
 	cli.IntFlag{
@@ -593,12 +598,7 @@ func setGlobalInternodeInterface(interfaceName string) {
 				ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
 				defer cancel()
 
-				lookupHost := globalDNSCache.LookupHost
-				if IsKubernetes() || IsDocker() {
-					lookupHost = net.DefaultResolver.LookupHost
-				}
-
-				haddrs, err := lookupHost(ctx, host)
+				haddrs, err := globalDNSCache.LookupHost(ctx, host)
 				if err == nil {
 					ip = haddrs[0]
 				}
@@ -636,12 +636,7 @@ func getServerListenAddrs() []string {
 		ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
 		defer cancel()
 
-		lookupHost := globalDNSCache.LookupHost
-		if IsKubernetes() || IsDocker() {
-			lookupHost = net.DefaultResolver.LookupHost
-		}
-
-		haddrs, err := lookupHost(ctx, host)
+		haddrs, err := globalDNSCache.LookupHost(ctx, host)
 		if err == nil {
 			for _, addr := range haddrs {
 				addrs.Add(net.JoinHostPort(addr, globalMinioPort))

cmd/utils.go

@@ -572,13 +572,8 @@ func ToS3ETag(etag string) string {
 
 // GetDefaultConnSettings returns default HTTP connection settings.
 func GetDefaultConnSettings() xhttp.ConnSettings {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = nil
-	}
-
 	return xhttp.ConnSettings{
-		LookupHost:  lookupHost,
+		LookupHost:  globalDNSCache.LookupHost,
 		DialTimeout: rest.DefaultTimeout,
 		RootCAs:     globalRootCAs,
 		TCPOptions:  globalTCPOptions,
@@ -588,13 +583,8 @@ func GetDefaultConnSettings() xhttp.ConnSettings {
 // NewInternodeHTTPTransport returns a transport for internode MinIO
 // connections.
 func NewInternodeHTTPTransport(maxIdleConnsPerHost int) func() http.RoundTripper {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = nil
-	}
-
 	return xhttp.ConnSettings{
-		LookupHost:       lookupHost,
+		LookupHost:       globalDNSCache.LookupHost,
 		DialTimeout:      rest.DefaultTimeout,
 		RootCAs:          globalRootCAs,
 		CipherSuites:     fips.TLSCiphers(),
@@ -607,13 +597,8 @@ func NewInternodeHTTPTransport(maxIdleConnsPerHost int) func() http.RoundTripper
 // NewCustomHTTPProxyTransport is used only for proxied requests, specifically
 // only supports HTTP/1.1
 func NewCustomHTTPProxyTransport() func() *http.Transport {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = nil
-	}
-
 	return xhttp.ConnSettings{
-		LookupHost:       lookupHost,
+		LookupHost:       globalDNSCache.LookupHost,
 		DialTimeout:      rest.DefaultTimeout,
 		RootCAs:          globalRootCAs,
 		CipherSuites:     fips.TLSCiphers(),
@@ -626,13 +611,8 @@ func NewCustomHTTPProxyTransport() func() *http.Transport {
 // NewHTTPTransportWithClientCerts returns a new http configuration
 // used while communicating with the cloud backends.
 func NewHTTPTransportWithClientCerts(clientCert, clientKey string) *http.Transport {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = nil
-	}
-
 	s := xhttp.ConnSettings{
-		LookupHost:  lookupHost,
+		LookupHost:  globalDNSCache.LookupHost,
 		DialTimeout: defaultDialTimeout,
 		RootCAs:     globalRootCAs,
 		TCPOptions:  globalTCPOptions,
@@ -663,14 +643,9 @@ const defaultDialTimeout = 5 * time.Second
 
 // NewHTTPTransportWithTimeout allows setting a timeout.
 func NewHTTPTransportWithTimeout(timeout time.Duration) *http.Transport {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = nil
-	}
-
 	return xhttp.ConnSettings{
 		DialContext: newCustomDialContext(),
-		LookupHost:  lookupHost,
+		LookupHost:  globalDNSCache.LookupHost,
 		DialTimeout: defaultDialTimeout,
 		RootCAs:     globalRootCAs,
 		TCPOptions:  globalTCPOptions,
@@ -702,14 +677,9 @@ func newCustomDialContext() xhttp.DialContext {
 // NewRemoteTargetHTTPTransport returns a new http configuration
 // used while communicating with the remote replication targets.
 func NewRemoteTargetHTTPTransport(insecure bool) func() *http.Transport {
-	lookupHost := globalDNSCache.LookupHost
-	if IsKubernetes() || IsDocker() {
-		lookupHost = nil
-	}
-
 	return xhttp.ConnSettings{
 		DialContext: newCustomDialContext(),
-		LookupHost:  lookupHost,
+		LookupHost:  globalDNSCache.LookupHost,
 		RootCAs:     globalRootCAs,
 		TCPOptions:  globalTCPOptions,
 		EnableHTTP2: false,