diff --git a/include/types/activity.h b/include/types/activity.h
index f58d759ca..9cf4e731d 100644
--- a/include/types/activity.h
+++ b/include/types/activity.h
@@ -51,6 +51,8 @@ struct activity {
 	struct freq_ctr cpust_1s;  // avg amount of half-ms stolen over last second
 	struct freq_ctr_period cpust_15s; // avg amount of half-ms stolen over last 15s
 	unsigned int avg_loop_us;  // average run time per loop over last 1024 runs
+	unsigned int accq_pushed;  // accept queue connections pushed
+	unsigned int accq_full;    // accept queue connection not pushed because full
 	char __pad[0]; // unused except to check remaining room
 	char __end[0] __attribute__((aligned(64))); // align size to 64.
 };
diff --git a/src/cli.c b/src/cli.c
index 3c2a55d75..ae5e4e8e3 100644
--- a/src/cli.c
+++ b/src/cli.c
@@ -1019,6 +1019,8 @@ static int cli_io_handler_show_activity(struct appctx *appctx)
 	chunk_appendf(&trash, "\ncpust_ms_1s:");  for (thr = 0; thr < global.nbthread; thr++) chunk_appendf(&trash, " %u", read_freq_ctr(&activity[thr].cpust_1s)/2);
 	chunk_appendf(&trash, "\ncpust_ms_15s:"); for (thr = 0; thr < global.nbthread; thr++) chunk_appendf(&trash, " %u", read_freq_ctr_period(&activity[thr].cpust_15s, 15000)/2);
 	chunk_appendf(&trash, "\navg_loop_us:");  for (thr = 0; thr < global.nbthread; thr++) chunk_appendf(&trash, " %u", swrate_avg(activity[thr].avg_loop_us, TIME_STATS_SAMPLES));
+	chunk_appendf(&trash, "\naccq_pushed:");  for (thr = 0; thr < global.nbthread; thr++) chunk_appendf(&trash, " %u", activity[thr].accq_pushed);
+	chunk_appendf(&trash, "\naccq_full:");    for (thr = 0; thr < global.nbthread; thr++) chunk_appendf(&trash, " %u", activity[thr].accq_full);
 
 	chunk_appendf(&trash, "\n");
 
diff --git a/src/listener.c b/src/listener.c
index 2b8c2dfab..0d7a1c480 100644
--- a/src/listener.c
+++ b/src/listener.c
@@ -852,13 +852,14 @@ void listener_accept(int fd)
 			 */
 			ring = &accept_queue_rings[t1];
 			if (accept_queue_push_mp(ring, cfd, l, &addr, laddr)) {
+				HA_ATOMIC_ADD(&activity[t1].accq_pushed, 1);
 				task_wakeup(ring->task, TASK_WOKEN_IO);
 				continue;
 			}
 			/* If the ring is full we do a synchronous accept on
 			 * the local thread here.
-			 * FIXME: we should update some stats here.
 			 */
+			HA_ATOMIC_ADD(&activity[t1].accq_full, 1);
 		}
 #endif // USE_THREAD