[#131] backoff strategy implemented

README.md

@@ -172,6 +172,13 @@ We can use this token for an entire hour, after that we will receive something l
  [{<<"reason">>,<<"ExpiredProviderToken">>}]}
 ```
 
+## Reconnection
+
+If network goes down or something unexpected happens the `gun` connection with APNs will go down. In that case `apns4erl` will send a message `{reconnecting, ServerPid}` to the client process, that means `apns4erl` lost the connection and it is trying to reconnect. Once the connection has been recover a `{connection_up, ServerPid}` message will be send.
+
+
+We implemented an *Exponential Backoff* strategy. We can set the *ceiling* time adding the `backoff_ceiling` variable on the `config` file. By default it is set to 10 (seconds).
+
 ## Close connections
 
 Apple recommends us to keep our connections open and avoid opening and closing very often. You can check the [Best Practices for Managing Connections](https://developer.apple.com/library/content/documentation/NetworkingInternet/Conceptual/RemoteNotificationsPG/CommunicatingwithAPNs.html) section.

rebar.lock

@@ -1,16 +1,16 @@
 {"1.1.0",
-[{<<"base64url">>,{pkg,<<"base64url">>,<<"0.0.1">>},0},
+[{<<"aleppo">>,{pkg,<<"inaka_aleppo">>,<<"1.0.0">>},3},
+ {<<"base64url">>,{pkg,<<"base64url">>,<<"0.0.1">>},0},
  {<<"cowlib">>,
   {git,"https://github.com/ninenines/cowlib",
        {ref,"0e7abe0b24593f131add272c275406d8ed231805"}},
   1},
- {<<"elvis_core">>,{pkg,<<"elvis_core">>,<<"0.3.2">>},1},
+ {<<"elvis">>,{pkg,<<"elvis_core">>,<<"0.3.2">>},1},
  {<<"goldrush">>,{pkg,<<"goldrush">>,<<"0.1.7">>},3},
  {<<"gun">>,
   {git,"https://github.com/ninenines/gun.git",
        {ref,"bc733a2ca5f7d07f997ad6edf184f775b23434aa"}},
   0},
- {<<"inaka_aleppo">>,{pkg,<<"inaka_aleppo">>,<<"1.0.0">>},3},
  {<<"jsx">>,{pkg,<<"jsx">>,<<"2.8.1">>},0},
  {<<"katana">>,{pkg,<<"katana">>,<<"0.3.1">>},0},
  {<<"katana_code">>,{pkg,<<"katana_code">>,<<"0.1.0">>},2},
@@ -22,10 +22,10 @@
  {<<"zipper">>,{pkg,<<"zipper">>,<<"1.0.0">>},2}]}.
 [
 {pkg_hash,[
+ {<<"aleppo">>, <<"8DB14CF16BB8C263C14FF4C3F69F64D7C849D40888944F4204D2CA74F1114CEB">>},
  {<<"base64url">>, <<"36A90125F5948E3AFD7BE97662A1504B934DD5DAC78451CA6E9ABF85A10286BE">>},
- {<<"elvis_core">>, <<"5424DBD17BEC4265A6808F49064CC1A8D2D1B593B5C4E94E747EDB157ABDE6DE">>},
+ {<<"elvis">>, <<"5424DBD17BEC4265A6808F49064CC1A8D2D1B593B5C4E94E747EDB157ABDE6DE">>},
  {<<"goldrush">>, <<"349A351D17C71C2FDAA18A6C2697562ABE136FEC945F147B381F0CF313160228">>},
- {<<"inaka_aleppo">>, <<"8DB14CF16BB8C263C14FF4C3F69F64D7C849D40888944F4204D2CA74F1114CEB">>},
  {<<"jsx">>, <<"1453B4EB3615ACB3E2CD0A105D27E6761E2ED2E501AC0B390F5BBEC497669846">>},
  {<<"katana">>, <<"7824B4F848E22088872780DD33C0CE6DADBCF2184200D4FA84D9040C9C55496E">>},
  {<<"katana_code">>, <<"C34F3926A258D6BEACD8D21F140F3D47D175501936431C460B144339D5271A0B">>},

src/apns_connection.erl

@@ -68,10 +68,11 @@
                          , type       := type()
                          }.
 
--type state()        :: #{ connection     := connection()
-                         , gun_connection := pid()
-                         , gun_monitor    := reference()
-                         , client         := pid()
+-type state()        :: #{ connection      := connection()
+                         , gun_connection  := pid()
+                         , client          := pid()
+                         , backoff         := non_neg_integer()
+                         , backoff_ceiling := non_neg_integer()
                          }.
 
 %%%===================================================================
@@ -146,18 +147,27 @@ push_notification(ConnectionName, Token, DeviceId, Notification, Headers) ->
                                   , Headers
                                   }).
 
+%% @doc Waits until receive the `connection_up` message
+-spec wait_apns_connection_up(pid()) -> {ok, pid()} | {error, timeout}.
+wait_apns_connection_up(Server) ->
+  receive
+    {connection_up, Server} -> {ok, Server};
+    {timeout, Server}       -> {error, timeout}
+  end.
+
 %%%===================================================================
 %%% gen_server callbacks
 %%%===================================================================
 
 -spec init({connection(), pid()}) -> {ok, State :: state(), timeout()}.
 init({Connection, Client}) ->
-  {GunMonitor, GunConnectionPid} = open_gun_connection(Connection),
+  GunConnectionPid = open_gun_connection(Connection),
 
-  {ok, #{ connection     => Connection
-        , gun_connection => GunConnectionPid
-        , gun_monitor    => GunMonitor
-        , client         => Client
+  {ok, #{ connection      => Connection
+        , gun_connection  => GunConnectionPid
+        , client          => Client
+        , backoff         => 1
+        , backoff_ceiling => application:get_env(apns, backoff_ceiling, 10)
         }, 0}.
 
 -spec handle_call( Request :: term(), From :: {pid(), term()}, State) ->
@@ -187,15 +197,37 @@ handle_cast(stop, State) ->
 handle_cast(_Request, State) ->
   {noreply, State}.
 
--spec handle_info(Info :: timeout() | term(), State) ->
-  {noreply, State}.
-handle_info( {'DOWN', GunMonitor, process, GunConnPid, _}
-           , #{ gun_connection := GunConnPid
-              , gun_monitor    := GunMonitor
-              , connection     := Connection
+-spec handle_info(Info :: timeout() | term(), State) -> {noreply, State}.
+handle_info( {gun_down, GunConn, http2, closed, _, _}
+           , #{ gun_connection  := GunConn
+              , client          := Client
+              , backoff         := Backoff
+              , backoff_ceiling := Ceiling
               } = State) ->
-  {GunMonitor2, GunConnPid2} = open_gun_connection(Connection),
-  {noreply, State#{gun_connection => GunConnPid2, gun_monitor => GunMonitor2}};
+  ok = gun:close(GunConn),
+  Client ! {reconnecting, self()},
+  Sleep = backoff(Backoff, Ceiling) * 1000, % seconds to wait before reconnect
+  {ok, _} = timer:send_after(Sleep, reconnect),
+  {noreply, State#{backoff => Backoff + 1}};
+handle_info(reconnect, State) ->
+  #{ connection      := Connection
+   , client          := Client
+   , backoff         := Backoff
+   , backoff_ceiling := Ceiling
+   } = State,
+  GunConn = open_gun_connection(Connection),
+  {ok, Timeout} = application:get_env(apns, timeout),
+  case gun:await_up(GunConn, Timeout) of
+    {ok, http2} ->
+      Client ! {connection_up, self()},
+      {noreply, State#{ gun_connection => GunConn
+                      , backoff        => 1}};
+    {error, timeout} ->
+      ok = gun:close(GunConn),
+      Sleep = backoff(Backoff, Ceiling) * 1000, % seconds to wait
+      {ok, _} = timer:send_after(Sleep, reconnect),
+      {noreply, State#{backoff => Backoff + 1}}
+  end;
 handle_info(timeout, #{gun_connection := GunConn, client := Client} = State) ->
   {ok, Timeout} = application:get_env(apns, timeout),
   case gun:await_up(GunConn, Timeout) of
@@ -254,9 +286,7 @@ type(#{type := Type}) ->
 %%% Internal Functions
 %%%===================================================================
 
--spec open_gun_connection(connection()) -> { GunMonitor       :: reference()
-                                           , GunConnectionPid :: pid()
-                                           }.
+-spec open_gun_connection(connection()) -> GunConnectionPid :: pid().
 open_gun_connection(Connection) ->
   Host = host(Connection),
   Port = port(Connection),
@@ -275,8 +305,7 @@ open_gun_connection(Connection) ->
                                    , #{ protocols => [http2]
                                       , transport_opts => TransportOpts
                                       }),
-  GunMonitor = monitor(process, GunConnectionPid),
-  {GunMonitor, GunConnectionPid}.
+  GunConnectionPid.
 
 -spec get_headers(apns:headers()) -> list().
 get_headers(Headers) ->
@@ -299,13 +328,6 @@ get_headers(Headers) ->
 get_device_path(DeviceId) ->
   <<"/3/device/", DeviceId/binary>>.
 
--spec wait_apns_connection_up(pid()) -> {ok, pid()} | {error, timeout}.
-wait_apns_connection_up(Server) ->
-  receive
-    {connection_up, Server} -> {ok, Server};
-    {timeout, Server}       -> {error, timeout}
-  end.
-
 -spec add_authorization_header(apns:headers(), apnd:token()) -> apns:headers().
 add_authorization_header(Headers, Token) ->
   Headers#{apns_auth_token => <<"bearer ", Token/binary>>}.
@@ -326,3 +348,13 @@ push(GunConn, DeviceId, HeadersMap, Notification) ->
       {Status, ResponseHeaders, DecodedBody};
     {error, timeout} -> timeout
   end.
+
+-spec backoff(non_neg_integer(), non_neg_integer()) -> non_neg_integer().
+backoff(N, Ceiling) ->
+  case (math:pow(2, N) - 1) of
+    R when R > Ceiling ->
+      Ceiling;
+    NextN ->
+      NString = float_to_list(NextN, [{decimals, 0}]),
+      list_to_integer(NString)
+  end.

test/connection_SUITE.erl

@@ -102,11 +102,14 @@ gun_connection_crashes(_Config) ->
   ok = mock_gun_open(),
   ok = mock_gun_await_up({ok, http2}),
   ConnectionName = my_connection2,
-  {ok, _ServerPid}  = apns:connect(cert, ConnectionName),
+  {ok, ServerPid}  = apns:connect(cert, ConnectionName),
   GunPid = apns_connection:gun_connection(ConnectionName),
   true = is_process_alive(GunPid),
-  GunPid ! crash,
+  GunPid ! {crash, ServerPid},
   ktn_task:wait_for(fun() -> is_process_alive(GunPid) end, false),
+  ktn_task:wait_for(fun() ->
+      apns_connection:gun_connection(ConnectionName) == GunPid
+    end, false),
   GunPid2 = apns_connection:gun_connection(ConnectionName),
   true = is_process_alive(GunPid2),
   true = (GunPid =/= GunPid2),
@@ -248,9 +251,9 @@ default_headers(_Config) ->
 -spec test_function() -> ok.
 test_function() ->
   receive
-    normal -> ok;
-    crash  -> exit(crashed);
-    _      -> test_function()
+    normal           -> ok;
+    {crash, Pid}     -> Pid ! {gun_down, self(), http2, closed, [], []};
+    _                -> test_function()
   end.
 
 -spec mock_gun_open() -> ok.