Tweak okay -> suspect config

Make it just a number of timeouts, without a new DWR being sent.

3 files changed, 42 insertions, 39 deletions

diff --git a/lib/diameter/doc/src/diameter.xml b/lib/diameter/doc/src/diameter.xml
index 75e95a9a3d..8e9ec06ff9 100644
--- a/lib/diameter/doc/src/diameter.xml
+++ b/lib/diameter/doc/src/diameter.xml
@@ -1132,10 +1132,10 @@ corresponding timeout (see below) or all fail.</p>
 Specifies configuration that alters the behaviour of the watchdog
 state machine.
 On key <c>okay</c>, the non-negative number of answered DWR
-messages required before transitioning from REOPEN to OKAY.
-On key <c>suspect</c>, the positive number of unanswered DWR messages
-before transitioning from OKAY to SUSPECT, or 0 to never make this
-transition.</p>
+messages before transitioning from REOPEN to OKAY.
+On key <c>suspect</c>, the number of watchdog timeouts before
+transitioning from OKAY to SUSPECT when DWR is unanswered, or 0 to
+not make the transition.</p>
 
 <p>
 Defaults to <c>[{okay, 3}, {suspect, 1}]</c>.
diff --git a/lib/diameter/src/base/diameter_watchdog.erl b/lib/diameter/src/base/diameter_watchdog.erl
index 0b32974efe..82ca603cf3 100644
--- a/lib/diameter/src/base/diameter_watchdog.erl
+++ b/lib/diameter/src/base/diameter_watchdog.erl
@@ -49,6 +49,8 @@
 
 -define(IS_NATURAL(N), (is_integer(N) andalso 0 =< N)).
 
+-define(CHOOSE(B,T,F), if (B) -> T; true -> F end).
+
 -record(config,
         {suspect = 1 :: non_neg_integer(),    %% OKAY -> SUSPECT
          okay    = 3 :: non_neg_integer()}).  %% REOPEN -> OKAY
@@ -61,7 +63,7 @@
                                 %% {M,F,A} -> integer() >= 0
          num_dwa = 0 :: -1 | non_neg_integer(),
                      %% number of DWAs received in reopen,
-                     %% or to send in okay before moving to suspect
+                     %% or number of timeouts before okay -> suspect
          %% end PCB
          parent = self() :: pid(),              %% service process
          transport       :: pid() | undefined,  %% peer_fsm process
@@ -424,7 +426,7 @@ transition({'DOWN', _, process, TPid, _Reason},
            #watchdog{transport = TPid,
                      status = T}
            = S) ->
-    set_watchdog(S#watchdog{status = case T of initial -> T; _ -> down end,
+    set_watchdog(S#watchdog{status = ?CHOOSE(initial == T, T, down),
                             pending = false,
                             transport = undefined});
 
@@ -668,9 +670,10 @@ timeout(#watchdog{status = okay,
     case N of
         1 ->
             S#watchdog{status = suspect};
-        _ ->  %% non-standard
-            send_watchdog(S#watchdog{pending = false,
-                                     num_dwa = decr(N)})
+        0 ->  %% non-standard: never move to suspect
+            S;
+        N ->  %% non-standard: more timeouts before moving
+            S#watchdog{num_dwa = N-1}
     end;
 
 %%   SUSPECT       Timer expires        CloseConnection()
@@ -725,11 +728,6 @@ timeout(#watchdog{status = T} = S)
        T == down ->
     restart(S).
 
-decr(0 = N) ->
-    N;
-decr(N) ->
-    N-1.
-
 %% restart/1
 
 restart(#watchdog{transport = undefined} = S) ->
diff --git a/lib/diameter/test/diameter_watchdog_SUITE.erl b/lib/diameter/test/diameter_watchdog_SUITE.erl
index 82244a1c7f..704bf110c7 100644
--- a/lib/diameter/test/diameter_watchdog_SUITE.erl
+++ b/lib/diameter/test/diameter_watchdog_SUITE.erl
@@ -89,16 +89,23 @@
 -define(INFO(T), #diameter_event{info = T}).
 
 %% Receive an event message from diameter.
--define(EVENT(T),
-        apply(fun() ->  %% apply to not bind T_
-                      receive #diameter_event{info = T = T_} ->
-                              log_event(T_)
-                      end
+-define(EVENT(T),    %% apply to not bind T_
+        apply(fun() ->
+                      receive ?INFO(T = T_) -> log_event(T_) end
               end,
               [])).
 
 %% Receive a watchdog event.
 -define(WD_EVENT(Ref), log_wd(element(4, ?EVENT({watchdog, Ref, _, _, _})))).
+-define(WD_EVENT(Ref, Ms),
+        apply(fun() ->
+                      receive ?INFO({watchdog, Ref, _, T_, _}) ->
+                              log_wd(T_)
+                      after Ms ->
+                              false
+                      end
+              end,
+              [])).
 
 %% Log to make failures identifiable.
 -define(LOG(T),     ?LOG("~p", [T])).
@@ -376,8 +383,8 @@ tpid(Ref, [[{ref, Ref},
 %% # suspect/1
 %% ===========================================================================
 
-%% Configure transports to require a set number of watchdogs before
-%% moving from OKAY to SUSPECT.
+%% Configure transports to require a set number of watchdog timeouts
+%% before moving from OKAY to SUSPECT.
 
 suspect(_) ->
     [] = run([[abuse, [suspect, N]] || N <- [0,1,3]]).
@@ -394,19 +401,21 @@ suspect(TRef, true, SvcName, _) ->
     {okay, _} = ?WD_EVENT(TRef);
 
 suspect(TRef, false, SvcName, 0) ->  %% SUSPECT disabled
-    %% Wait 2+ watchdogs and see that two unanswered watchdogs have
-    %% been sent.
-    [2,0,0,0] = receive
-                    ?INFO({watchdog, TRef, _, _, _} = T) -> T
-                after 28000 ->
-                        wd_counts(SvcName)
-                end;
+    %% Wait 2+ watchdogs and see that only one watchdog has been sent.
+    false = ?WD_EVENT(TRef, 28000),
+    [1,0,0,0] = wd_counts(SvcName);
 
 suspect(TRef, false, SvcName, N) ->
-    {okay, suspect} = ?WD_EVENT(TRef),
-    [N,0,0,0] = wd_counts(SvcName),
-    {suspect, down} = ?WD_EVENT(TRef),
-    [N,0,0,0] = wd_counts(SvcName).
+    %% Check that no watchdog transition takes place within N+
+    %% watchdogs ...
+    false = ?WD_EVENT(TRef, N*10000+8000),
+    [1,0,0,0] = wd_counts(SvcName),
+    %% ... but that the connection then becomes suspect ...
+    {okay, suspect} = ?WD_EVENT(TRef, 10000),
+    [1,0,0,0] = wd_counts(SvcName),
+    %% ... and goes down.
+    {suspect, down} = ?WD_EVENT(TRef, 18000),
+    [1,0,0,0] = wd_counts(SvcName).
 
 %% abuse/1
 
@@ -470,13 +479,9 @@ ok(TRef, SvcName, Down, 0) ->
     %% Connection comes up without watchdog exchange.
     {Down, okay} = ?WD_EVENT(TRef),
     [1,0,0,0] = wd_counts(SvcName),
-    %% Wait 2+ watchdog timeout to see that the connection stays up and
-    %% two watchdogs are exchanged.
-    ok = receive ?INFO({watchdog, TRef, _, _, _} = T) ->
-                 T
-         after 28000 ->
-                 ok
-         end,
+    %% Wait 2+ watchdog timeouts to see that the connection stays up
+    %% and two watchdogs are exchanged.
+    false = ?WD_EVENT(TRef, 28000),
     [3,0,0,2] = wd_counts(SvcName);
 
 ok(TRef, SvcName, Down, N) ->