riak_test/tests/repl_fs_stat_caching.erl
Engel A. Sanchez 556cb7210c Ensure riak_repl service is up across the board
Trying to use the repl features before newly started nodes have
riak_repl completely initialized leads to all sorts of nasty crashes and
noise. Frequently it makes fullsync stuck forever, which makes a lot of
the tests fail.

This also tweaks the AAE fullsync tests to remove assumptions about
failure stats when AAE transient errors occur. The behavior in the
handling of those errors has changed recently with the introduction of
soft exits.
2014-12-18 16:07:00 -05:00

110 lines
3.4 KiB
Erlang

%% @doc Tests to ensure a stalling or blocking fssource process does not
%% cause status call to timeout. Useful for only 2.0 and up (and up is
%% a regression test).
-module(repl_fs_stat_caching).
-behavior(riak_test).
-include_lib("eunit/include/eunit.hrl").
-define(TEST_BUCKET, <<"repl_fs_stat_caching">>).
-export([confirm/0]).
confirm() ->
{{SrcLead, SrcCluster}, {SinkLead, _SinkCluster}} = setup(),
SinkPort = repl_util:get_cluster_mgr_port(SinkLead),
repl_util:connect_cluster(SrcLead, "127.0.0.1", SinkPort),
lager:info("Loading source cluster"),
[] = repl_util:do_write(SrcLead, 1, 1000, ?TEST_BUCKET, 1),
repl_util:enable_fullsync(SrcLead, "sink"),
rpc:call(SrcLead, riak_repl_console, fullsync, [["start", "sink"]]),
% and now, the actual test.
% find a random fssource, suspend it, and then ensure we can get a
% status.
{ok, Suspended} = suspend_an_fs_source(SrcCluster),
lager:info("Suspended: ~p", [Suspended]),
{ok, Status} = rt:riak_repl(SrcLead, "status"),
FailLine = "RPC to '" ++ atom_to_list(SrcLead) ++ "' failed: timeout\n",
?assertNotEqual(FailLine, Status),
true = rpc:block_call(node(Suspended), erlang, resume_process, [Suspended]),
pass.
setup() ->
rt:set_conf(all, [{"buckets.default.allow_mult", "false"}]),
NodeCount = rt_config:get(num_nodes, 6),
lager:info("Deploy ~p nodes", [NodeCount]),
Nodes = rt:deploy_nodes(NodeCount, cluster_conf(), [riak_kv, riak_repl]),
SplitSize = NodeCount div 2,
{SourceNodes, SinkNodes} = lists:split(SplitSize, Nodes),
lager:info("making cluster Source from ~p", [SourceNodes]),
repl_util:make_cluster(SourceNodes),
lager:info("making cluster Sink from ~p", [SinkNodes]),
repl_util:make_cluster(SinkNodes),
SrcHead = hd(SourceNodes),
SinkHead = hd(SinkNodes),
repl_util:name_cluster(SrcHead, "source"),
repl_util:name_cluster(SinkHead, "sink"),
rt:wait_until_ring_converged(SourceNodes),
rt:wait_until_ring_converged(SinkNodes),
rt:wait_until_transfers_complete(SourceNodes),
rt:wait_until_transfers_complete(SinkNodes),
ok = repl_util:wait_until_leader_converge(SourceNodes),
ok = repl_util:wait_until_leader_converge(SinkNodes),
SourceLead = repl_util:get_leader(SrcHead),
SinkLead = repl_util:get_leader(SinkHead),
{{SourceLead, SourceNodes}, {SinkLead, SinkNodes}}.
cluster_conf() ->
[
{riak_repl, [
{fullsync_on_connect, false},
{fullsync_interval, disabled},
{max_fssource_cluster, 3},
{max_fssource_node, 1},
{max_fssink_node, 20},
{rtq_max_bytes, 1048576}
]}
].
suspend_an_fs_source([]) ->
{error, no_nodes};
suspend_an_fs_source(Nodes) ->
suspend_an_fs_source(Nodes, 10000).
suspend_an_fs_source([_Node | _Tail], 0) ->
{error, tries_ran_out};
suspend_an_fs_source([Node | Tail], TriesLeft) ->
Pids = rpc:call(Node, riak_repl2_fssource_sup, enabled, []),
case maybe_suspend_an_fs_source(Node, Pids) of
false ->
suspend_an_fs_source(Tail ++ [Node], TriesLeft - 1);
Pid ->
{ok, Pid}
end.
maybe_suspend_an_fs_source(_Node, []) ->
false;
maybe_suspend_an_fs_source(Node, [{_Remote, Pid} | Tail]) ->
case rpc:block_call(Node, erlang, suspend_process, [Pid]) of
false ->
maybe_suspend_an_fs_source(Node, Tail);
true ->
Pid
end.