Merge pull request #101 from safing/fix/dns-resolving-resilience

Improve dns resolving resilience

resolver/config.go

@@ -29,11 +29,11 @@ var (
 
 		// Cloudflare (encrypted DNS, with malware protection)
 		`dot://1.1.1.2:853?verify=cloudflare-dns.com&name=Cloudflare&blockedif=zeroip`,
-		`dot://1.0.0.2:853?verify=cloudflare-dns.com&name=Cloudflare&blockedif=zeroip`,
+		// `dot://1.0.0.2:853?verify=cloudflare-dns.com&name=Cloudflare&blockedif=zeroip`,
 
 		// AdGuard (encrypted DNS, default flavor)
-		`dot://176.103.130.130:853?verify=dns.adguard.com&name=AdGuard&blockedif=zeroip`,
+		// `dot://176.103.130.130:853?verify=dns.adguard.com&name=AdGuard&blockedif=zeroip`,
-		`dot://176.103.130.131:853?verify=dns.adguard.com&name=AdGuard&blockedif=zeroip`,
+		// `dot://176.103.130.131:853?verify=dns.adguard.com&name=AdGuard&blockedif=zeroip`,
 
 		// Foundation for Applied Privacy (encrypted DNS)
 		// `dot://94.130.106.88:853?verify=dot1.applied-privacy.net&name=AppliedPrivacy`,

resolver/resolve.go

@@ -45,6 +45,12 @@ var (
 	ErrNoCompliance = fmt.Errorf("%w: no compliant resolvers for this query", ErrBlocked)
 )
 
+const (
+	minTTL     = 60           // 1 Minute
+	minMDnsTTL = 60           // 1 Minute
+	maxTTL     = 24 * 60 * 60 // 24 hours
+)
+
 // BlockedUpstreamError is returned when a DNS request
 // has been blocked by the upstream server.
 type BlockedUpstreamError struct {
@@ -296,18 +302,16 @@ resolveLoop:
 					// we are offline and this is not an online check query
 					return nil, ErrOffline
 				default:
+					// includes ErrTimeout
 					log.Tracer(ctx).Debugf("resolver: failed to resolve %s: %s", q.FQDN, err)
 				}
-			} else {
+			}
-				// no error
 			if rrCache == nil {
-					// defensive: assume NXDomain
+				continue
-					return nil, ErrNotFound
 			}
 			break resolveLoop
 		}
 	}
-	}
 
 	// check for error
 	if err != nil {
@@ -326,7 +330,7 @@ resolveLoop:
 	// cache if enabled
 	if !q.NoCaching {
 		// persist to database
-		rrCache.Clean(600)
+		rrCache.Clean(minTTL)
 		err = rrCache.Save()
 		if err != nil {
 			log.Warningf("resolver: failed to cache RR for %s%s: %s", q.FQDN, q.QType.String(), err)

resolver/resolver-mdns.go

@@ -276,7 +276,7 @@ func handleMDNSMessages(ctx context.Context, messages chan *dns.Msg) error {
 
 			var questionID string
 			if saveFullRequest {
-				rrCache.Clean(60)
+				rrCache.Clean(minMDnsTTL)
 				err := rrCache.Save()
 				if err != nil {
 					log.Warningf("resolver: failed to cache RR %s: %s", rrCache.Domain, err)
@@ -304,7 +304,7 @@ func handleMDNSMessages(ctx context.Context, messages chan *dns.Msg) error {
 					Server:      mDNSResolver.Server,
 					ServerScope: mDNSResolver.ServerIPScope,
 				}
-				rrCache.Clean(60)
+				rrCache.Clean(minMDnsTTL)
 				err := rrCache.Save()
 				if err != nil {
 					log.Warningf("resolver: failed to cache RR %s: %s", rrCache.Domain, err)

resolver/resolver-tcp.go

@@ -419,7 +419,7 @@ func (mgr *tcpResolverConnMgr) handleQueryResponse(conn *dns.Conn, msg *dns.Msg)
 
 	// persist to database
 	rrCache := inFlight.MakeCacheRecord(msg)
-	rrCache.Clean(600)
+	rrCache.Clean(minTTL)
 	err := rrCache.Save()
 	if err != nil {
 		log.Warningf(

resolver/resolver.go

@@ -28,7 +28,7 @@ const (
 
 var (
 	// FailThreshold is amount of errors a resolvers must experience in order to be regarded as failed.
-	FailThreshold = 5
+	FailThreshold = 20
 )
 
 // Resolver holds information about an active resolver.

resolver/rrcache.go

@@ -72,9 +72,12 @@ func (rrCache *RRCache) Clean(minExpires uint32) {
 		header.Ttl = 17
 	}
 
-	// TTL must be at least minExpires
+	// TTL range limits
-	if lowestTTL < minExpires {
+	switch {
+	case lowestTTL < minExpires:
 		lowestTTL = minExpires
+	case lowestTTL > maxTTL:
+		lowestTTL = maxTTL
 	}
 
 	// shorten caching