Keep a per-DNS-server moving average of query latency.

4 files changed, 22 insertions, 8 deletions

diff --git a/src/cache.c b/src/cache.c
index 689451c..437e777 100644
--- a/src/cache.c
+++ b/src/cache.c
@@ -1664,19 +1664,18 @@ int cache_make_stat(struct txt_record *t)
 	  {
 	    char *new, *lenp;
 	    int port, newlen, bytes_avail, bytes_needed;
-	    unsigned int queries = 0, failed_queries = 0, nxdomain_replies = 0;
+	    unsigned int queries = 0, failed_queries = 0;
 	    for (serv1 = serv; serv1; serv1 = serv1->next)
 	      if (!(serv1->flags & SERV_MARK) && sockaddr_isequal(&serv->addr, &serv1->addr))
 		{
 		  serv1->flags |= SERV_MARK;
 		  queries += serv1->queries;
 		  failed_queries += serv1->failed_queries;
-		  nxdomain_replies += serv1->nxdomain_replies;
 		}
 	    port = prettyprint_addr(&serv->addr, daemon->addrbuff);
 	    lenp = p++; /* length */
 	    bytes_avail = bufflen - (p - buff );
-	    bytes_needed = snprintf(p, bytes_avail, "%s#%d %u %u %u", daemon->addrbuff, port, queries, failed_queries, nxdomain_replies);
+	    bytes_needed = snprintf(p, bytes_avail, "%s#%d %u %u", daemon->addrbuff, port, queries, failed_queries);
 	    if (bytes_needed >= bytes_avail)
 	      {
 		/* expand buffer if necessary */
@@ -1747,6 +1746,8 @@ void dump_cache(time_t now)
       {
 	int port;
 	unsigned int queries = 0, failed_queries = 0, nxdomain_replies = 0;
+	unsigned int sigma_latency = 0, count_latency = 0;
+
 	for (serv1 = serv; serv1; serv1 = serv1->next)
 	  if (!(serv1->flags & SERV_MARK) && sockaddr_isequal(&serv->addr, &serv1->addr))
 	    {
@@ -1754,10 +1755,12 @@ void dump_cache(time_t now)
 	      queries += serv1->queries;
 	      failed_queries += serv1->failed_queries;
 	      nxdomain_replies += serv1->nxdomain_replies;
+	      sigma_latency += serv1->query_latency;
+	      count_latency++;
 	    }
 	port = prettyprint_addr(&serv->addr, daemon->addrbuff);
-	my_syslog(LOG_INFO, _("server %s#%d: queries sent %u, retried or failed %u, nxdomain replies %u"),
-		  daemon->addrbuff, port, queries, failed_queries, nxdomain_replies);
+	my_syslog(LOG_INFO, _("server %s#%d: queries sent %u, retried or failed %u, nxdomain replies %u, avg. latency %ums"),
+		  daemon->addrbuff, port, queries, failed_queries, nxdomain_replies, sigma_latency/count_latency);
       }
 
   if (option_bool(OPT_DEBUG) || option_bool(OPT_LOG))
diff --git a/src/dbus.c b/src/dbus.c
index 782fdc9..ebd8d7f 100644
--- a/src/dbus.c
+++ b/src/dbus.c
@@ -682,6 +682,8 @@ static DBusMessage *dbus_get_server_metrics(DBusMessage* message)
       {
 	unsigned int port;
 	unsigned int queries = 0, failed_queries = 0, nxdomain_replies = 0;
+	unsigned int sigma_latency = 0, count_latency = 0;
+	
 	struct server *serv1;
 
 	for (serv1 = serv; serv1; serv1 = serv1->next)
@@ -691,6 +693,8 @@ static DBusMessage *dbus_get_server_metrics(DBusMessage* message)
 	      queries += serv1->queries;
 	      failed_queries += serv1->failed_queries;
 	      nxdomain_replies += serv1->nxdomain_replies;
+	      sigma_latency += serv1->query_latency;
+	      count_latency++;
 	    }
 	
 	dbus_message_iter_open_container(&server_array, DBUS_TYPE_ARRAY, "{ss}", &dict_array);
@@ -702,7 +706,8 @@ static DBusMessage *dbus_get_server_metrics(DBusMessage* message)
 	add_dict_int(&dict_array, "queries", serv->queries);
 	add_dict_int(&dict_array, "failed_queries", serv->failed_queries);
 	add_dict_int(&dict_array, "nxdomain", serv->nxdomain_replies);
-	
+	add_dict_int(&dict_array, "latency", sigma_latency/count_latency;
+		     
 	dbus_message_iter_close_container(&server_array, &dict_array);
       }
   
diff --git a/src/dnsmasq.h b/src/dnsmasq.h
index ef412b5..f8d7c55 100644
--- a/src/dnsmasq.h
+++ b/src/dnsmasq.h
@@ -587,6 +587,7 @@ struct server {
   int tcpfd, edns_pktsz;
   time_t pktsz_reduced;
   unsigned int queries, failed_queries, nxdomain_replies;
+  unsigned int query_latency, mma_latency;
   time_t forwardtime;
   int forwardcount;
 #ifdef HAVE_LOOP
diff --git a/src/forward.c b/src/forward.c
index 2134528..e1ec30a 100644
--- a/src/forward.c
+++ b/src/forward.c
@@ -553,8 +553,7 @@ static int forward_query(int udpfd, union mysockaddr *udpaddr,
   
   if (forwarded || is_dnssec)
     {
-      if (daemon->fast_retry_time != 0)
-	forward->forward_timestamp = dnsmasq_milliseconds();
+      forward->forward_timestamp = dnsmasq_milliseconds();
       return 1;
     }
   
@@ -1237,6 +1236,12 @@ void reply_query(int fd, time_t now)
      answers, to conserve file descriptors, and to save work reading and
      discarding answers for other upstreams. */
   free_rfds(&forward->rfds);
+
+  /* calculate modified moving average of server latency */
+  server->mma_latency += dnsmasq_milliseconds() - forward->forward_timestamp - server->query_latency;
+  /* denominator controls how many queries we average over. */
+  server->query_latency = server->mma_latency/128;
+  
   
 #ifdef HAVE_DNSSEC
   if ((forward->sentto->flags & SERV_DO_DNSSEC) &&