Re-enable disk_monitor in case of parser failures

Parser failures could be transient on start-up, so retry a few times
before giving up.

rabbitmq-server#1178
[#143558437]

3 files changed, 67 insertions, 12 deletions

diff --git a/Makefile b/Makefile
index c824002b64..e40145dd1c 100644
--- a/Makefile
+++ b/Makefile
@@ -107,7 +107,9 @@ define PROJECT_ENV
 	    {queue_explicit_gc_run_operation_threshold, 1000},
 	    {lazy_queue_explicit_gc_run_operation_threshold, 1000},
 	    {background_gc_enabled, false},
-	    {background_gc_target_interval, 60000}
+	    {background_gc_target_interval, 60000},
+	    {disk_monitor_enable_retries, 10},
+	    {disk_monitor_enable_interval, 120000}
 	  ]
 endef
 
diff --git a/src/rabbit_disk_monitor.erl b/src/rabbit_disk_monitor.erl
index b2548cb61a..86f9b4016e 100644
--- a/src/rabbit_disk_monitor.erl
+++ b/src/rabbit_disk_monitor.erl
@@ -65,7 +65,12 @@
           alarmed,
           %% is monitoring enabled? false on unsupported
           %% platforms
-          enabled
+          enabled,
+          %% number of retries to enable monitoring if it fails
+          %% on start-up
+          retries,
+          %% Interval between retries
+          interval
 }).
 
 %%----------------------------------------------------------------------------
@@ -114,20 +119,17 @@ start_link(Args) ->
 
 init([Limit]) ->
     Dir = dir(),
+    {ok, Retries} = application:get_env(rabbit, disk_monitor_enable_retries),
+    {ok, Interval} = application:get_env(rabbit, disk_monitor_enable_interval),
     State = #state{dir          = Dir,
                    min_interval = ?DEFAULT_MIN_DISK_CHECK_INTERVAL,
                    max_interval = ?DEFAULT_MAX_DISK_CHECK_INTERVAL,
                    alarmed      = false,
-                   enabled      = true},
-    case {catch get_disk_free(Dir),
-          vm_memory_monitor:get_total_memory()} of
-        {N1, N2} when is_integer(N1), is_integer(N2) ->
-            {ok, start_timer(set_disk_limits(State, Limit))};
-        Err ->
-            rabbit_log:info("Disabling disk free space monitoring "
-                            "on unsupported platform:~n~p~n", [Err]),
-            {ok, State#state{enabled = false}}
-    end.
+                   enabled      = true,
+                   limit        = Limit,
+                   retries      = Retries,
+                   interval     = Interval},
+    {ok, enable(State)}.
 
 handle_call(get_disk_free_limit, _From, State = #state{limit = Limit}) ->
     {reply, Limit, State};
@@ -161,6 +163,8 @@ handle_call(_Request, _From, State) ->
 handle_cast(_Request, State) ->
     {noreply, State}.
 
+handle_info(try_enable, #state{retries = Retries} = State) ->
+    {noreply, enable(State#state{retries = Retries - 1})};
 handle_info(update, State) ->
     {noreply, start_timer(internal_update(State))};
 
@@ -261,3 +265,20 @@ interval(#state{limit        = Limit,
                 max_interval = MaxInterval}) ->
     IdealInterval = 2 * (Actual - Limit) / ?FAST_RATE,
     trunc(erlang:max(MinInterval, erlang:min(MaxInterval, IdealInterval))).
+
+enable(#state{retries = 0} = State) ->
+    State;
+enable(#state{dir = Dir, interval = Interval, limit = Limit, retries = Retries}
+       = State) ->
+    case {catch get_disk_free(Dir),
+          vm_memory_monitor:get_total_memory()} of
+        {N1, N2} when is_integer(N1), is_integer(N2) ->
+            rabbit_log:info("Enabling disk free space monitoring~n", []),
+            start_timer(set_disk_limits(State, Limit));
+        Err ->
+            rabbit_log:info("Disabling disk free space monitoring "
+                            "on unsupported platform, ~p retries left:~n~p~n",
+                            [Retries, Err]),
+            timer:send_after(Interval, self(), try_enable),
+            State#state{enabled = false}
+    end.
diff --git a/test/unit_inbroker_non_parallel_SUITE.erl b/test/unit_inbroker_non_parallel_SUITE.erl
index 266f0d30c0..68b7a15a8a 100644
--- a/test/unit_inbroker_non_parallel_SUITE.erl
+++ b/test/unit_inbroker_non_parallel_SUITE.erl
@@ -35,6 +35,7 @@ groups() ->
           app_management, %% Restart RabbitMQ.
           channel_statistics, %% Expect specific statistics.
           disk_monitor, %% Replace rabbit_misc module.
+          disk_monitor_enable,
           file_handle_cache, %% Change FHC limit.
           head_message_timestamp_statistics, %% Expect specific statistics.
           log_management, %% Check log files.
@@ -744,6 +745,37 @@ disk_monitor1(_Config) ->
     meck:unload(rabbit_misc),
     passed.
 
+disk_monitor_enable(Config) ->
+    passed = rabbit_ct_broker_helpers:rpc(Config, 0,
+      ?MODULE, disk_monitor_enable1, [Config]).
+
+disk_monitor_enable1(_Config) ->
+    case os:type() of
+        {unix, _} ->
+            disk_monitor_enable1();
+        _ ->
+            %% skip windows testing
+            skipped
+    end.
+
+disk_monitor_enable1() ->
+    ok = meck:new(rabbit_misc, [passthrough]),
+    ok = meck:expect(rabbit_misc, os_cmd, fun(_) -> "\n" end),
+    application:set_env(rabbit, disk_monitor_enable_retries, 20000),
+    application:set_env(rabbit, disk_monitor_enable_interval, 100),
+    ok = rabbit_sup:stop_child(rabbit_disk_monitor_sup),
+    ok = rabbit_sup:start_delayed_restartable_child(rabbit_disk_monitor, [1000]),
+    undefined = rabbit_disk_monitor:get_disk_free(),
+    Cmd = "Filesystem 1024-blocks      Used Available Capacity  iused     ifree %iused  Mounted on\n/dev/disk1   975798272 234783364 740758908    25% 58759839 185189727   24%   /\n",
+    ok = meck:expect(rabbit_misc, os_cmd, fun(_) -> Cmd end),
+    timer:sleep(1000),
+    Bytes = 740758908 * 1024,
+    Bytes = rabbit_disk_monitor:get_disk_free(),
+    meck:unload(rabbit_misc),
+    application:set_env(rabbit, disk_monitor_enable_retries, 10),
+    application:set_env(rabbit, disk_monitor_enable_interval, 120000),
+    passed.
+
 %% ---------------------------------------------------------------------------
 %% rabbitmqctl helpers.
 %% ---------------------------------------------------------------------------