riak_test/tests/repl_handoff_deadlock_common.erl
Doug Rohrer cad958481d Test deadlock between Repl Fullsync and Handoff. Requires riak_repl and riak_kv fixes:
Tests that provoke deadlock causing full_sync to hang when handoff causes vnode to exit before the fold request is started.
2015-04-24 15:23:06 -04:00

152 lines
5.5 KiB
Erlang

%% -------------------------------------------------------------------
%%
%% Copyright (c) 2015 Basho Technologies, Inc.
%%
%% This file is provided to you under the Apache License,
%% Version 2.0 (the "License"); you may not use this file
%% except in compliance with the License. You may obtain
%% a copy of the License at
%%
%% http://www.apache.org/licenses/LICENSE-2.0
%%
%% Unless required by applicable law or agreed to in writing,
%% software distributed under the License is distributed on an
%% "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
%% KIND, either express or implied. See the License for the
%% specific language governing permissions and limitations
%% under the License.
%%
%% -------------------------------------------------------------------
%%% @copyright (C) 2015, Basho Technologies
%%% @doc
%%% riak_test for timing deadlock on riak_repl2_keylist_server:diff_bloom and riak_repl_aae_source:finish_sending_differences
%%% which can occur when the owner VNode is handed off after the original monitor (MonRef) is created. Neither function
%%% handles the {'DOWN', MonRef, process, VNodePid, normal} case (which happens when the node exits after handoff.
%%% Note that this uses code from verify_counter_repl as its tests, as that seemed to be able to evoke the issue, even without intercepts
%%%
%%% Also tests fixes for riak_repl2_fscoordinator which used to cache the owner of an index at replication start, but those could change with handoff.
%%% @end
-module(repl_handoff_deadlock_common).
-export([confirm/1]).
-include_lib("eunit/include/eunit.hrl").
-define(BUCKET, <<"counter-bucket">>).
-define(KEY, <<"counter-key">>).
-define(CONF(Strategy), [
{riak_kv,
[
%% Specify fast building of AAE trees
{anti_entropy, {on, []}},
{anti_entropy_build_limit, {100, 1000}},
{anti_entropy_concurrency, 100}
]
},
{riak_repl,
[
{fullsync_strategy, Strategy},
{fullsync_on_connect, false},
{fullsync_interval, disabled},
%% Force usage of Bloom to invoke race
{fullsync_direct_percentage_limit, 0},
{fullsync_direct_limit, 1}
]}
]).
confirm(Strategy) ->
inets:start(),
lager:info("Testing fullsync handoff deadlock with strategy ~p~n", [Strategy]),
{ClusterA, ClusterB} = make_clusters(Strategy),
%% Simulate stop of 1/10th of vnodes before fold begins to provoke deadlock
[
rt_intercept:add(Node, {riak_core_vnode_master,
[{{command_return_vnode, 4},
stop_vnode_after_bloom_fold_request_succeeds}]})
|| {_, Node} <- ClusterA ++ ClusterB],
%% Write the data to both sides of the cluster
write_data(hd(ClusterA), 1, 1000),
write_data(hd(ClusterB), 1001, 1000),
%% let the repl flow
repl_power_activate(ClusterA, ClusterB),
verify_data(hd(ClusterA), 2000),
verify_data(hd(ClusterB), 2000),
pass.
make_clusters(KeylistOrAae) ->
Nodes = rt:deploy_nodes(6, ?CONF(KeylistOrAae), [riak_kv, riak_repl]),
{ClusterA, ClusterB} = lists:split(3, Nodes),
A = make_cluster(ClusterA, "A"),
B = make_cluster(ClusterB, "B"),
{A, B}.
make_cluster(Nodes, Name) ->
repl_util:make_cluster(Nodes),
repl_util:name_cluster(hd(Nodes), Name),
repl_util:wait_until_leader_converge(Nodes),
Clients = [ rt:pbc(Node) || Node <- Nodes ],
lists:zip(Clients, Nodes).
write_data({Client, _Node}, Start, Count) ->
[riakc_pb_socket:put(Client, riakc_obj:new(?BUCKET, integer_to_binary(Num), integer_to_binary(Num)))
|| Num <- lists:seq(Start, (Start-1) + Count)].
verify_data({Client, Node}, Count) ->
[
begin
case (riakc_pb_socket:get(Client, ?BUCKET, integer_to_binary(Num), [{notfound_ok, false}])) of
{error, notfound} ->
erlang:error({not_found, lists:flatten(io_lib:format("Could not find ~p in cluster with node ~p", [Num, Node]))});
{ok, Object} ->
?assertEqual(riakc_obj:get_value(Object), integer_to_binary(Num))
end
end
|| Num <- lists:seq(1, Count)].
%% Set up bi-directional full sync replication.
repl_power_activate(ClusterA, ClusterB) ->
lager:info("repl power...ACTIVATE!"),
LeaderA = get_leader(hd(ClusterA)),
info("got leader A"),
LeaderB = get_leader(hd(ClusterB)),
info("Got leader B"),
MgrPortA = get_mgr_port(hd(ClusterA)),
info("Got manager port A"),
MgrPortB = get_mgr_port(hd(ClusterB)),
info("Got manager port B"),
info("connecting A to B"),
repl_util:connect_cluster(LeaderA, "127.0.0.1", MgrPortB),
?assertEqual(ok, repl_util:wait_for_connection(LeaderA, "B")),
info("A connected to B"),
info("connecting B to A"),
repl_util:connect_cluster(LeaderB, "127.0.0.1", MgrPortA),
?assertEqual(ok, repl_util:wait_for_connection(LeaderB, "A")),
info("B connected to A"),
info("Enabling Fullsync bi-directional"),
repl_util:enable_fullsync(LeaderA, "B"),
info("Enabled A->B"),
repl_util:enable_fullsync(LeaderB, "A"),
info("Enabled B->A"),
info("Awaiting fullsync completion"),
repl_util:start_and_wait_until_fullsync_complete(LeaderA),
info("A->B complete"),
repl_util:start_and_wait_until_fullsync_complete(LeaderB),
info("B->A complete").
get_leader({_, Node}) ->
rpc:call(Node, riak_core_cluster_mgr, get_leader, []).
get_mgr_port({_, Node}) ->
{ok, {_IP, Port}} = rpc:call(Node, application, get_env,
[riak_core, cluster_mgr]),
Port.
info(Message) ->
lager:info(Message).