Use supervisors to manage crashes better.

Now crashing the FSM that manages the backend will cause the
connections to drop instead of just be orphaned.

closes dustin#12

Change-Id: I99105f8ef9abf8949049ae0e9ea63cfe117f7cdc

etc/couchdb/local.d/mccouch.ini

@@ -1,3 +1,3 @@
 ; Add this to etc/couchdb/local.d
 [daemons]
-mc_daemon={mc_daemon, start_link, ["kv", true]}
+mc_daemon={mc_sup, start_link, ["kv", true]}

src/mc_conn_sup.erl

@@ -0,0 +1,26 @@
+-module(mc_conn_sup).
+
+-behaviour(supervisor).
+
+-export([start_link/0]).
+
+-export([init/1]).
+
+-export([start_connection/1]).
+
+-define(SERVER, ?MODULE).
+
+start_link() ->
+    supervisor:start_link({local, ?SERVER}, ?MODULE, []).
+
+init([]) ->
+    {ok, {{simple_one_for_one, 0, 1},
+          [{mc_connection, {mc_connection, start_link, [mc_daemon]},
+            temporary, brutal_kill, worker, [mc_connection]}]}}.
+
+start_connection(NS) ->
+    {ok, Pid} = supervisor:start_child(?MODULE, [NS]),
+    gen_tcp:controlling_process(NS, Pid),
+    %% Tell this mc_connection it's the controlling process and ready to go.
+    Pid ! go,
+    {ok, Pid}.

src/mc_connection.erl

@@ -1,10 +1,13 @@
 -module (mc_connection).
 
--export([loop/2]).
+-export([start_link/2, init/2]).
 -export([respond/5, respond/4]).
 
 -include("mc_constants.hrl").
 
+start_link(Handler, Socket) ->
+    {ok, spawn_link(?MODULE, init, [Socket, Handler])}.
+
 bin_size(undefined) -> 0;
 bin_size(IoList) -> iolist_size(IoList).
 
@@ -90,6 +93,11 @@ process_message(Socket, StorageServer, {ok, <<?REQ_MAGIC:8, OpCode:8, KeyLen:16,
         Res -> respond(Socket, OpCode, Opaque, Res)
     end.
 
+init(Socket, Handler) ->
+    %% The spawner will tell us when we are the controlling process.
+    receive go -> ok end,
+    loop(Socket, Handler).
+
 loop(Socket, Handler) ->
     process_message(Socket, Handler, gen_tcp:recv(Socket, ?HEADER_LEN)),
     loop(Socket, Handler).

src/mc_daemon.erl

@@ -22,16 +22,15 @@
 -include("couch_db.hrl").
 -include("mc_constants.hrl").
 
--record(state, {mc_serv, db, json_mode, setqs=0, terminal_opaque=0}).
+-record(state, {db, json_mode, setqs=0, terminal_opaque=0}).
 
 start_link(DbName, JsonMode) ->
     gen_fsm:start_link({local, ?SERVER}, ?MODULE, [DbName, JsonMode], []).
 
 init([DbName, JsonMode]) ->
     ?LOG_INFO("MC daemon: starting: json_mode=~p.", [JsonMode]),
-    {ok, S} = mc_tcp_listener:start_link(11213, self()),
     {ok, processing,
-     #state{mc_serv=S, db=list_to_binary(DbName), json_mode=JsonMode}}.
+     #state{db=list_to_binary(DbName), json_mode=JsonMode}}.
 
 db_name(VBucket, State)->
     iolist_to_binary([State#state.db, $/, integer_to_list(VBucket)]).

src/mc_sup.erl

@@ -0,0 +1,48 @@
+-module(mc_sup).
+
+-behaviour(supervisor).
+
+-export([start_link/2]).
+
+-export([init/1]).
+
+-define(SERVER, ?MODULE).
+
+start_link(DbName, JsonMode) ->
+    application:start(sasl),
+    supervisor:start_link({local, ?SERVER}, ?MODULE, [DbName, JsonMode]).
+
+%%--------------------------------------------------------------------
+%% @private
+%% @doc
+%% Whenever a supervisor is started using supervisor:start_link/[2,3],
+%% this function is called by the new process to find out about
+%% restart strategy, maximum restart frequency and child
+%% specifications.
+%%
+%% @spec init(Args) -> {ok, {SupFlags, [ChildSpec]}} |
+%%                     ignore |
+%%                     {error, Reason}
+%% @end
+%%--------------------------------------------------------------------
+init([DbName, JsonMode]) ->
+    RestartStrategy = one_for_all,
+    MaxRestarts = 1000,
+    MaxSecondsBetweenRestarts = 3600,
+
+    SupFlags = {RestartStrategy, MaxRestarts, MaxSecondsBetweenRestarts},
+
+    Restart = permanent,
+    Shutdown = 2000,
+    Type = worker,
+
+    Daemon = {mc_daemon, {mc_daemon, start_link, [DbName, JsonMode]},
+              Restart, Shutdown, Type, [mc_daemon]},
+
+    TcpListener = {mc_tcp_listener, {mc_tcp_listener, start_link, [11213, mc_daemon]},
+                   Restart, Shutdown, Type, [mc_tcp_listener, mc_daemon]},
+
+    ConnSup = {mc_conn_sup, {mc_conn_sup, start_link, []},
+               Restart, Shutdown, supervisor, dynamic},
+
+    {ok, {SupFlags, [Daemon, TcpListener, ConnSup]}}.

src/mc_tcp_listener.erl

@@ -31,6 +31,5 @@ init(PortNum, StorageServer) ->
 % Accept incoming connections
 accept_loop(LS, StorageServer) ->
     {ok, NS} = gen_tcp:accept(LS),
-    Pid = spawn(mc_connection, loop, [NS, StorageServer]),
-    gen_tcp:controlling_process(NS, Pid),
+    mc_conn_sup:start_connection(NS),
     accept_loop(LS, StorageServer).