Merge remote-tracking branch 'basho/develop-2.2' into bch-merge-with-2.2

README.md

@@ -181,8 +181,8 @@ to tell riak_test about them. The method of choice is to create a
     {rtdev_path, [{root,     "/home/you/rt/riak"},
                   {current,  "/home/you/rt/riak/current"},
                   {previous, "/home/you/rt/riak/riak-2.0.6"},
-                  {legacy,   "/home/you/rt/riak/riak-1.4.12"}
+                  {legacy,   "/home/you/rt/riak/riak-1.4.12"},
-                  {'2.0.2',   "/home/you/rt/riak/riak-2.0.2"}
+                  {'2.0.2',   "/home/you/rt/riak/riak-2.0.2"},
                   {'2.0.4',   "/home/you/rt/riak/riak-2.0.4"}
                  ]}
 ]}.

intercepts/intercept.erl

@@ -22,7 +22,7 @@
 %% Export explicit API but also send compile directive to export all
 %% because some of these private functions are useful in their own
 %% right.
--export([add/3, add/4]).
+-export([add/3, add/4, clean/1]).
 -compile(export_all).
 
 -type abstract_code() :: term().
@@ -54,7 +54,7 @@
 %%             functions.
 %%
 %%             E.g. `[{{update_perform,2}, sleep_update_perform}]'
--spec add(module(), module(), mapping()) -> ok.
+-spec add(module(), module(), mapping(), string()) -> ok.
 add(Target, Intercept, Mapping, OutDir) ->
     Original = ?ORIGINAL(Target),
     TargetAC = get_abstract_code(Target),
@@ -66,9 +66,22 @@ add(Target, Intercept, Mapping, OutDir) ->
     ok = compile_and_load(Original, OrigAC, OutDir),
     ok = compile_and_load(Target, ProxyAC, OutDir).
 
+-spec add(module(), module(), mapping()) -> ok.
 add(Target, Intercept, Mapping) ->
     add(Target, Intercept, Mapping, undefined).
 
+%% @doc Cleanup proxy and backuped original module
+-spec clean(module()) -> ok|{error, term()}.
+clean(Target) ->
+    _ = code:purge(Target),
+    _ = code:purge(?ORIGINAL(Target)),
+    case code:load_file(Target) of
+        {module, Target} ->
+            ok;
+        {error, Reason} ->
+            {error, Reason}
+    end.
+
 %% @private
 %%
 %% @doc Compile the abstract code `AC' and load it into the code server.

intercepts/riak_kv_vnode_intercepts.erl

@@ -159,3 +159,9 @@ corrupting_handle_handoff_data(BinObj0, State) ->
 
 corrupt_binary(O) ->
     crypto:rand_bytes(byte_size(O)).
+
+put_as_readrepair(Preflist, BKey, Obj, ReqId, StartTime, Options) ->
+    ?M:put_orig(Preflist, BKey, Obj, ReqId, StartTime, [rr | Options]).
+
+coord_put_as_readrepair(Preflist, BKey, Obj, ReqId, StartTime, Options) ->
+    ?M:coord_put_orig(Preflist, BKey, Obj, ReqId, StartTime, [rr | Options]).

src/rt_intercept.erl

@@ -78,6 +78,12 @@ add(Node, {Target, Intercept, Mapping}, OutDir) ->
     NMapping = [transform_anon_fun(M) || M <- Mapping],
     ok = rpc:call(Node, intercept, add, [Target, Intercept, NMapping, OutDir]).
 
+clean(Node, Targets) when is_list(Targets) ->
+    [ok = clean(Node, T) || T <- Targets],
+    ok;
+clean(Node, Target) ->
+    ok = rpc:call(Node, intercept, clean, [Target]).
+
 %% The following function transforms anonymous function mappings passed
 %% from an Erlang shell. Anonymous intercept functions from compiled code
 %% require the developer to supply free variables themselves, and also

tests/repl_process_leak.erl

@@ -0,0 +1,143 @@
+%% @doc The purpose of thie test is to ensure the realtime helpers on both
+%% the source and sink sides properly exit when a connection is flakey; ie
+%% then there are errors and not out-right closes of the connection.
+
+-module(repl_process_leak).
+-behavior(riak_test).
+-export([confirm/0]).
+-include_lib("eunit/include/eunit.hrl").
+
+-define(SEND_ERROR_INTERVAL, 500).
+
+confirm() ->
+    Conf = [
+        {riak_repl, [
+            {fullsync_on_connect, false},
+            {fullsync_interval, disabled}
+        ]}
+    ],
+
+    lager:info("deploying 2 nodes"),
+    Nodes = rt:deploy_nodes(2, Conf, [riak_kv, riak_repl]),
+
+    [SourceNode, SinkNode] = Nodes,
+
+    lager:info("nameing clusters"),
+    repl_util:name_cluster(SourceNode, "source"),
+    repl_util:name_cluster(SinkNode, "sink"),
+
+    {ok, {_IP, Port}} = rpc:call(SinkNode, application, get_env, [riak_core, cluster_mgr]),
+
+    lager:info("connecting clusters using port ~p", [Port]),
+    repl_util:connect_cluster(SourceNode, "127.0.0.1", Port),
+    repl_util:wait_for_connection(SourceNode, "sink"),
+
+    lager:info("enabling and starting realtime"),
+    repl_util:enable_realtime(SourceNode, "sink"),
+    repl_util:start_realtime(SourceNode, "sink"),
+
+    lager:info("testing for leaks on flakey sink"),
+    flakey_sink(SourceNode, SinkNode),
+
+    lager:info("testing for leaks on flakey source"),
+    flakey_source(SourceNode, SinkNode),
+
+    pass.
+
+flakey_sink(_SourceNode, SinkNode) ->
+    InitialCount = rpc:call(SinkNode, erlang, system_info, [process_count]),
+    ProcCounts = send_sink_tcp_errors(SinkNode, 20, [InitialCount]),
+
+    Smallest = lists:min(ProcCounts),
+    Biggest = lists:max(ProcCounts),
+    ?assert(2 =< Biggest - Smallest),
+    %?assertEqual(InitialProcCount, PostProcCount),
+    % the process count is increasing, but the helper did die
+    true.
+
+send_sink_tcp_errors(_SinkNode, 0, Acc) ->
+    Acc;
+
+send_sink_tcp_errors(SinkNode, N, Acc) ->
+    case rpc:call(SinkNode, riak_repl2_rtsink_conn_sup, started, []) of
+        [] ->
+            timer:sleep(?SEND_ERROR_INTERVAL),
+            send_sink_tcp_errors(SinkNode, N, Acc);
+        [P | _] ->
+            SysStatus = sys:get_status(P),
+            {status, P, _Modul, [_PDict, _Status, _, _, Data]} = SysStatus,
+            [_Header, _Data1, Data2] = Data,
+            {data, [{"State", StateRec}]} = Data2,
+            [Helper | _] = lists:filter(fun(E) ->
+                is_pid(E)
+            end, tuple_to_list(StateRec)),
+            HelpMon = erlang:monitor(process, Helper),
+            P ! {tcp_error, <<>>, test},
+            Mon = erlang:monitor(process, P),
+            receive {'DOWN', Mon, process, P, _} -> ok end,
+            receive
+                {'DOWN', HelpMon, process, Helper, _} ->
+                    ok
+                after 10000 ->
+                    throw("helper didn't die")
+            end,
+            timer:sleep(?SEND_ERROR_INTERVAL),
+            Procs = rpc:call(SinkNode, erlang, system_info, [process_count]),
+            send_sink_tcp_errors(SinkNode, N - 1, [Procs | Acc])
+    end.
+
+flakey_source(SourceNode, _SinkNode) ->
+    InitialProcCount = rpc:call(SourceNode, erlang, system_info, [process_count]),
+    ProcCounts = send_source_tcp_errors(SourceNode, 20, [InitialProcCount]),
+
+    Biggest = lists:max(ProcCounts),
+    Smallest = lists:min(ProcCounts),
+    %lager:info("initial: ~p; post: ~p", [InitialProcCount, PostProcCount]),
+    %?assertEqual(InitialProcCount, PostProcCount).
+    ?assert(2 =< Biggest - Smallest),
+    true.
+
+send_source_tcp_errors(_SourceNode, 0, Acc) ->
+    Acc;
+
+send_source_tcp_errors(SourceNode, N, Acc) ->
+    List = rpc:call(SourceNode, riak_repl2_rtsource_conn_sup, enabled, []),
+    case proplists:get_value("sink", List) of
+        undefined ->
+            timer:sleep(?SEND_ERROR_INTERVAL),
+            send_source_tcp_errors(SourceNode, N, Acc);
+        Pid ->
+            lager:debug("Get the status"),
+            SysStatus = try sys:get_status(Pid) of
+                S -> S
+            catch
+                W:Y ->
+                    lager:info("Sys failed due to ~p:~p", [W,Y]),
+                    {status, Pid, undefined, [undefined, undefined, undefined, undefined, [undefined, undefined, {data, [{"State", {Pid}}]}]]}
+            end,
+            {status, Pid, _Module, [_PDict, _Status, _, _, Data]} = SysStatus,
+            [_Header, _Data1, Data2] = Data,
+            {data, [{"State", StateRec}]} = Data2,
+            [Helper | _] = lists:filter(fun(E) ->
+                is_pid(E)
+            end, tuple_to_list(StateRec)),
+            lager:debug("mon the hlepr"),
+            HelperMon = erlang:monitor(process, Helper),
+            lager:debug("Send the murder"),
+            Pid ! {tcp_error, <<>>, test},
+            Mon = erlang:monitor(process, Pid),
+            lager:debug("Wait for deaths"),
+            receive
+                {'DOWN', Mon, process, Pid, _} -> ok
+            end,
+            receive
+                {'DOWN', HelperMon, process, Helper, _} ->
+                    ok
+                after 10000 ->
+                    throw("Helper didn't die")
+            end,
+            timer:sleep(?SEND_ERROR_INTERVAL),
+            Count = rpc:call(SourceNode, erlang, system_info, [process_count]),
+            send_source_tcp_errors(SourceNode, N - 1, [Count | Acc])
+    end.
+

tests/repl_rt_heartbeat.erl

@@ -65,8 +65,14 @@ confirm() ->
     suspend_heartbeat_messages(LeaderA),
 
     %% sleep longer than the HB timeout interval to force re-connection;
-    %% and give it time to restart the RT connection. Wait an extra 2 seconds.
+    %% and give it time to restart the RT connection.
-    timer:sleep(timer:seconds(?HB_TIMEOUT) + 2000),
+    %% Since it's possible we may disable heartbeats right after a heartbeat has been fired,
+    %% it can take up to 2*?HB_TIMEOUT seconds to detect a missed heartbeat. The extra second
+    %% is to avoid rare race conditions due to the timeouts lining up exactly. Not the prettiest
+    %% solution, but it failed so rarely at 2*HB_TIMEOUT, that this should be good enough
+    %% in practice, and it beats having to write a bunch of fancy intercepts to verify that
+    %% the timeout has been hit internally.
+    timer:sleep(timer:seconds(?HB_TIMEOUT*2) + 1000),
 
     %% Verify that RT connection has restarted by noting that it's Pid has changed
     RTConnPid2 = get_rt_conn_pid(LeaderA),

tests/verify_2i_eqc.erl

@@ -73,6 +73,7 @@
 -behaviour(riak_test).
 -export([confirm/0]).
 
+-define(NEG_LIMIT, -1000000). %% Due to sext encoding bug, do not permit negative numbers less than this to be generated.
 -define(MAX_CLUSTER_SIZE, 1).
 -define(MAX_FIELDS, 1).
 -define(FIELDS, ["i" ++ integer_to_list(N) || N <- lists:seq(1, ?MAX_FIELDS)]).
@@ -113,6 +114,7 @@ confirm() ->
     TestingTime = rt_config:get(eqc_testing_time, 120),
     lager:info("Will run in cluster of size ~p for ~p seconds.",
                [Size, TestingTime]),
+    rt:set_backend(eleveldb),
     Nodes = rt:build_cluster(Size),
     ?assert(eqc:quickcheck(
               eqc:testing_time(TestingTime, ?MODULE:prop_test(Nodes)))),
@@ -171,7 +173,24 @@ gen_term() ->
 
 %% Generates, with equal likelihood, either a smallish or a largish integer.
 gen_int_term() ->
-    oneof([int(), largeint()]).
+    oneof([int(), gen_large_int()]).
+
+%% XXX FIXME
+%% Ensure that very large negative numbers are not used as values.
+%%
+%% This is because the version of sext used in riak has an encoding
+%% bug for these values.  Until we can fix this, this test needs to
+%% deal with and work around the sext encoding bug.
+%%
+%% When that bug is fixed, this commit should be reverted/code
+%% deleted.
+gen_large_int() ->
+    ?LET(I, largeint(), no_large_neg_ints(I)).
+
+no_large_neg_ints(I) when I < ?NEG_LIMIT ->
+    abs(I);
+no_large_neg_ints(I) ->
+    I.
 
 %% Generates a random binary.
 gen_bin_term() ->

tests/verify_object_limits.erl

@@ -36,6 +36,7 @@
 confirm() ->
     [Node1] = rt:build_cluster(1, [{riak_kv, [
                         {ring_creation_size, 8},
+                        {anti_entropy, {off,[]}},
                         {max_object_size, ?MAX_SIZE},
                         {warn_object_size, ?WARN_SIZE},
                         {max_siblings, ?MAX_SIBLINGS},
@@ -51,6 +52,9 @@ confirm() ->
                                                 [{allow_mult, true}])),
     verify_size_limits(C, Node1),
     verify_sibling_limits(C, Node1),
+    lager:notice("Starting readrepair section of test"),
+    verify_readrepair_ignore_max_size(C, Node1),
+    verify_readrepair_ignore_max_sib(C, Node1),
     pass.
 
 verify_size_limits(C, Node1) ->
@@ -128,3 +132,47 @@ verify_sibling_limits(C, Node1) ->
     lager:info("Result when too many siblings : ~p", [Res]),
     ?assertMatch({error,_},  Res),
     ok.
+
+verify_readrepair_ignore_max_size(C, Node1) ->
+    % Add intercept to treat all vnode puts as readrepairs
+    Intercept = {riak_kv_vnode, [{{put, 6}, put_as_readrepair},{{coord_put,6}, coord_put_as_readrepair}]},
+    ok = rt_intercept:add(Node1, Intercept),
+    % Do put with value greater than max size and confirm warning
+    lager:info("Checking readrepair put of size ~p, expecting ok result and log warning", [?MAX_SIZE*2]),
+    K = <<"rrsizetest">>,
+    V = <<0:(?MAX_SIZE*2)/integer-unit:8>>,
+    O = riakc_obj:new(?BUCKET, K, V),
+    ?assertMatch(ok, riakc_pb_socket:put(C, O)),
+    verify_size_write_warning(Node1, K, ?MAX_SIZE*2),
+    % Clean intercept
+    ok = rt_intercept:clean(Node1, riak_kv_vnode),
+    ok.
+
+verify_readrepair_ignore_max_sib(C, Node1) ->
+    lager:info("Checking sibling warning on readrepair above max siblings=~p", [?MAX_SIBLINGS]),
+    K = <<"rrsibtest">>,
+    V = <<"sibtest">>,
+    O = riakc_obj:new(?BUCKET, K, V),
+    % Force sibling error
+    [?assertMatch(ok, riakc_pb_socket:put(C, O)) 
+     || _ <- lists:seq(1, ?MAX_SIBLINGS)],
+    Res = riakc_pb_socket:put(C, O),
+    lager:info("Result when too many siblings : ~p", [Res]),
+    ?assertMatch({error,_},  Res),
+    % Add intercept to spoof writes as readrepair
+    Intercept = {riak_kv_vnode, [{{put, 6}, put_as_readrepair},{{coord_put,6}, coord_put_as_readrepair}]},
+    ok = rt_intercept:add(Node1, Intercept),
+    % Verify readrepair writes return ok and log warning
+    lager:info("Verifying succesful put above max_siblings with readrepair"),
+    ?assertMatch(ok, riakc_pb_socket:put(C, O)),
+    P = io_lib:format("warning.*siblings.*~p.*~p.*(~p)",
+                      [?BUCKET, K, ?MAX_SIBLINGS+1]),
+    Found = rt:expect_in_log(Node1, P),
+    lager:info("Looking for sibling warning: ~p", [Found]),
+    ?assertEqual(true, Found),
+    % Clean intercept
+    ok = rt_intercept:clean(Node1, riak_kv_vnode),
+    ok.
+
+
+