syn_test_suite_helper.erl 13 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380
  1. %% ==========================================================================================================
  2. %% Syn - A global Process Registry and Process Group manager.
  3. %%
  4. %% The MIT License (MIT)
  5. %%
  6. %% Copyright (c) 2015-2021 Roberto Ostinelli <roberto@ostinelli.net> and Neato Robotics, Inc.
  7. %%
  8. %% Permission is hereby granted, free of charge, to any person obtaining a copy
  9. %% of this software and associated documentation files (the "Software"), to deal
  10. %% in the Software without restriction, including without limitation the rights
  11. %% to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  12. %% copies of the Software, and to permit persons to whom the Software is
  13. %% furnished to do so, subject to the following conditions:
  14. %%
  15. %% The above copyright notice and this permission notice shall be included in
  16. %% all copies or substantial portions of the Software.
  17. %%
  18. %% THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  19. %% IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  20. %% FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  21. %% AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  22. %% LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  23. %% OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  24. %% THE SOFTWARE.
  25. %% ==========================================================================================================
  26. -module(syn_test_suite_helper).
  27. %% API
  28. -export([init_cluster/1, end_cluster/2]).
  29. -export([start_slave/1, stop_slave/1]).
  30. -export([connect_node/1, disconnect_node/1]).
  31. -export([clean_after_test/0]).
  32. -export([start_process/0, start_process/1, start_process/2]).
  33. -export([kill_process/1]).
  34. -export([wait_cluster_mesh_connected/1]).
  35. -export([wait_process_name_ready/1, wait_process_name_ready/2]).
  36. -export([wait_message_queue_empty/0]).
  37. -export([assert_cluster/2]).
  38. -export([assert_registry_scope_subcluster/3, assert_pg_scope_subcluster/3]).
  39. -export([assert_received_messages/1]).
  40. -export([assert_empty_queue/0]).
  41. -export([assert_wait/2]).
  42. -export([send_error_logger_to_disk/0]).
  43. %% internal
  44. -export([process_main/0]).
  45. %% macro
  46. -define(DEFAULT_WAIT_TIMEOUT, 5000).
  47. -define(UNEXPECTED_MESSAGES_WAIT_TIMEOUT, 1000).
  48. %% ===================================================================
  49. %% API
  50. %% ===================================================================
  51. init_cluster(NodesCount) ->
  52. SlavesCount = NodesCount - 1,
  53. {Nodes, NodesConfig} = lists:foldl(fun(I, {AccNodes, AccNodesConfig}) ->
  54. IBin = integer_to_binary(I),
  55. NodeShortName = list_to_atom(binary_to_list(<<"syn_slave_", IBin/binary>>)),
  56. {ok, SlaveNode} = start_slave(NodeShortName),
  57. %% connect
  58. lists:foreach(fun(N) ->
  59. rpc:call(SlaveNode, syn_test_suite_helper, connect_node, [N])
  60. end, AccNodes),
  61. %% config
  62. {
  63. [SlaveNode | AccNodes],
  64. [{NodeShortName, SlaveNode} | AccNodesConfig]
  65. }
  66. end, {[], []}, lists:seq(1, SlavesCount)),
  67. %% wait full cluster
  68. case syn_test_suite_helper:wait_cluster_mesh_connected([node()] ++ Nodes) of
  69. ok ->
  70. %% config
  71. NodesConfig;
  72. Other ->
  73. ct:pal("*********** Could not get full cluster of ~p nodes, skipping", [NodesCount]),
  74. {error_initializing_cluster, Other}
  75. end.
  76. end_cluster(NodesCount, Config) ->
  77. SlavesCount = NodesCount - 1,
  78. %% clean
  79. clean_after_test(),
  80. %% shutdown
  81. lists:foreach(fun(I) ->
  82. IBin = integer_to_binary(I),
  83. NodeShortName = list_to_atom(binary_to_list(<<"syn_slave_", IBin/binary>>)),
  84. SlaveNode = proplists:get_value(NodeShortName, Config),
  85. connect_node(SlaveNode),
  86. stop_slave(NodeShortName)
  87. end, lists:seq(1, SlavesCount)),
  88. %% wait
  89. timer:sleep(1000).
  90. start_slave(NodeShortName) ->
  91. %% start slave
  92. {ok, Node} = ct_slave:start(NodeShortName, [
  93. {boot_timeout, 10},
  94. {erl_flags, "-connect_all false -kernel dist_auto_connect never"}
  95. ]),
  96. %% add syn code path to slaves
  97. CodePath = lists:filter(fun(Path) ->
  98. nomatch =/= string:find(Path, "/syn/")
  99. end, code:get_path()),
  100. true = rpc:call(Node, code, set_path, [CodePath]),
  101. %% return
  102. {ok, Node}.
  103. stop_slave(NodeShortName) ->
  104. {ok, _} = ct_slave:stop(NodeShortName).
  105. connect_node(Node) ->
  106. net_kernel:connect_node(Node).
  107. disconnect_node(Node) ->
  108. erlang:disconnect_node(Node).
  109. clean_after_test() ->
  110. Nodes = [node() | nodes()],
  111. %% shutdown
  112. lists:foreach(fun(Node) ->
  113. %% close syn
  114. rpc:call(Node, application, stop, [syn]),
  115. %% clean env
  116. rpc:call(Node, application, unset_env, [syn, event_handler]),
  117. rpc:call(Node, application, unset_env, [syn, strict_mode]),
  118. %% messages
  119. flush_inbox()
  120. end, Nodes).
  121. start_process() ->
  122. Pid = spawn(fun process_main/0),
  123. Pid.
  124. start_process(Node) when is_atom(Node) ->
  125. Pid = spawn(Node, fun process_main/0),
  126. Pid;
  127. start_process(Loop) when is_function(Loop) ->
  128. Pid = spawn(Loop),
  129. Pid.
  130. start_process(Node, Loop) ->
  131. Pid = spawn(Node, Loop),
  132. Pid.
  133. kill_process(RegisteredName) when is_atom(RegisteredName) ->
  134. case whereis(RegisteredName) of
  135. undefined -> ok;
  136. Pid -> kill_process(Pid)
  137. end;
  138. kill_process(Pid) when is_pid(Pid) ->
  139. case rpc:call(node(Pid), erlang, is_process_alive, [Pid]) of
  140. true ->
  141. MRef = monitor(process, Pid),
  142. exit(Pid, kill),
  143. receive
  144. {'DOWN', MRef, process, Pid, _Reason} -> ok
  145. after ?DEFAULT_WAIT_TIMEOUT ->
  146. ct:fail("~n\tCould not kill process ~p~n", [Pid])
  147. end;
  148. false ->
  149. ok
  150. end.
  151. wait_cluster_mesh_connected(Nodes) ->
  152. wait_cluster_mesh_connected(Nodes, os:system_time(millisecond)).
  153. wait_cluster_mesh_connected(Nodes, StartAt) ->
  154. AllSynced = lists:all(fun(Node) ->
  155. RemoteNodes = rpc:call(Node, erlang, nodes, []),
  156. AllNodes = [Node | RemoteNodes],
  157. lists:sort(AllNodes) == lists:sort(Nodes)
  158. end, Nodes),
  159. case AllSynced of
  160. true ->
  161. ok;
  162. false ->
  163. case os:system_time(millisecond) - StartAt > ?DEFAULT_WAIT_TIMEOUT of
  164. true ->
  165. {error, {could_not_init_cluster, Nodes}};
  166. false ->
  167. timer:sleep(50),
  168. wait_cluster_mesh_connected(Nodes, StartAt)
  169. end
  170. end.
  171. wait_process_name_ready(Name) ->
  172. wait_process_name_ready(Name, os:system_time(millisecond)).
  173. wait_process_name_ready(Name, StartAt) ->
  174. timer:sleep(50),
  175. case whereis(Name) of
  176. undefined ->
  177. case os:system_time(millisecond) - StartAt > ?DEFAULT_WAIT_TIMEOUT of
  178. true ->
  179. ct:fail("~n\tProcess with name ~p didn't come alive~n", [Name]);
  180. false ->
  181. wait_process_name_ready(Name, StartAt)
  182. end;
  183. Pid ->
  184. case process_info(Pid, status) of
  185. {status, waiting} ->
  186. ok;
  187. Other ->
  188. case os:system_time(millisecond) - StartAt > ?DEFAULT_WAIT_TIMEOUT of
  189. true ->
  190. ct:fail("~n\tProcess with name ~p didn't come ready~n\tStatus: ~p~n", [Name, Other]);
  191. false ->
  192. wait_process_name_ready(Name, StartAt)
  193. end
  194. end
  195. end.
  196. wait_message_queue_empty() ->
  197. timer:sleep(500),
  198. syn_test_suite_helper:assert_wait(
  199. ok,
  200. fun() ->
  201. flush_inbox(),
  202. syn_test_suite_helper:assert_empty_queue(self())
  203. end
  204. ).
  205. assert_cluster(Node, ExpectedNodes) ->
  206. assert_cluster(Node, ExpectedNodes, os:system_time(millisecond)).
  207. assert_cluster(Node, ExpectedNodes, StartAt) ->
  208. Nodes = rpc:call(Node, erlang, nodes, []),
  209. case do_assert_cluster(Nodes, ExpectedNodes, StartAt) of
  210. continue -> assert_cluster(Node, ExpectedNodes, StartAt);
  211. _ -> ok
  212. end.
  213. assert_registry_scope_subcluster(Node, Scope, ExpectedNodes) ->
  214. do_assert_scope_subcluster(registry, Node, Scope, ExpectedNodes).
  215. assert_pg_scope_subcluster(Node, Scope, ExpectedNodes) ->
  216. do_assert_scope_subcluster(pg, Node, Scope, ExpectedNodes).
  217. assert_received_messages(Messages) ->
  218. assert_received_messages(Messages, []).
  219. assert_received_messages([], UnexpectedMessages) ->
  220. assert_received_messages_wait([], UnexpectedMessages);
  221. assert_received_messages(Messages, UnexpectedMessages) ->
  222. receive
  223. Message ->
  224. case lists:member(Message, Messages) of
  225. true ->
  226. Messages1 = lists:delete(Message, Messages),
  227. assert_received_messages(Messages1, UnexpectedMessages);
  228. false ->
  229. assert_received_messages(Messages, [Message | UnexpectedMessages])
  230. end
  231. after ?DEFAULT_WAIT_TIMEOUT ->
  232. assert_received_messages_evaluate(Messages, UnexpectedMessages)
  233. end.
  234. assert_received_messages_wait(MissingMessages, UnexpectedMessages) ->
  235. receive
  236. Message ->
  237. assert_received_messages_wait(MissingMessages, [Message | UnexpectedMessages])
  238. after ?UNEXPECTED_MESSAGES_WAIT_TIMEOUT ->
  239. assert_received_messages_evaluate(MissingMessages, UnexpectedMessages)
  240. end.
  241. assert_received_messages_evaluate([], []) ->
  242. ok;
  243. assert_received_messages_evaluate(MissingMessages, UnexpectedMessages) ->
  244. ct:fail("~n\tReceive messages error (line ~p)~n\tMissing: ~p~n\tUnexpected: ~p~n",
  245. [get_line_from_stacktrace(), lists:reverse(MissingMessages), lists:reverse(UnexpectedMessages)]
  246. ).
  247. assert_empty_queue() ->
  248. assert_empty_queue([]).
  249. assert_empty_queue(UnexpectedMessages) ->
  250. receive
  251. Message ->
  252. assert_empty_queue([Message | UnexpectedMessages])
  253. after ?UNEXPECTED_MESSAGES_WAIT_TIMEOUT ->
  254. case UnexpectedMessages of
  255. [] -> ok;
  256. _ -> ct:fail("~n\tMessage queue was not empty, got:~n\t~p~n", [UnexpectedMessages])
  257. end
  258. end.
  259. assert_wait(ExpectedResult, Fun) ->
  260. assert_wait(ExpectedResult, Fun, os:system_time(millisecond)).
  261. assert_wait(ExpectedResult, Fun, StartAt) ->
  262. case Fun() of
  263. ExpectedResult ->
  264. ok;
  265. Result ->
  266. case os:system_time(millisecond) - StartAt > ?DEFAULT_WAIT_TIMEOUT of
  267. true ->
  268. ct:fail("~n\tExpected: ~p~n\tActual: ~p~n", [ExpectedResult, Result]);
  269. false ->
  270. timer:sleep(50),
  271. assert_wait(ExpectedResult, Fun, StartAt)
  272. end
  273. end.
  274. send_error_logger_to_disk() ->
  275. error_logger:logfile({open, atom_to_list(node())}).
  276. %% ===================================================================
  277. %% Internal
  278. %% ===================================================================
  279. process_main() ->
  280. receive
  281. _ ->
  282. process_main()
  283. end.
  284. do_assert_scope_subcluster(Type, Node, Scope, ExpectedNodes) ->
  285. do_assert_scope_subcluster(Type, Node, Scope, ExpectedNodes, os:system_time(millisecond)).
  286. do_assert_scope_subcluster(Type, Node, Scope, ExpectedNodes, StartAt) ->
  287. Nodes = rpc:call(Node, syn, subcluster_nodes, [Type, Scope]),
  288. case do_assert_cluster(Nodes, ExpectedNodes, StartAt) of
  289. continue -> do_assert_scope_subcluster(Type, Node, Scope, ExpectedNodes, StartAt);
  290. _ -> ok
  291. end.
  292. do_assert_cluster(Nodes, ExpectedNodes, StartAt) ->
  293. ExpectedCount = length(ExpectedNodes),
  294. %% count nodes
  295. case length(Nodes) of
  296. ExpectedCount ->
  297. %% loop nodes
  298. RemainingNodes = lists:filter(fun(N) -> not lists:member(N, ExpectedNodes) end, Nodes),
  299. case length(RemainingNodes) of
  300. 0 ->
  301. ok;
  302. _ ->
  303. case os:system_time(millisecond) - StartAt > ?DEFAULT_WAIT_TIMEOUT of
  304. true ->
  305. ct:fail("~n\tInvalid subcluster~n\tExpected: ~p~n\tActual: ~p~n\tLine: ~p~n",
  306. [ExpectedNodes, Nodes, get_line_from_stacktrace()]
  307. );
  308. false ->
  309. timer:sleep(50),
  310. continue
  311. end
  312. end;
  313. _ ->
  314. case os:system_time(millisecond) - StartAt > ?DEFAULT_WAIT_TIMEOUT of
  315. true ->
  316. ct:fail("~n\tInvalid subcluster~n\tExpected: ~p~n\tActual: ~p~n\tLine: ~p~n",
  317. [ExpectedNodes, Nodes, get_line_from_stacktrace()]
  318. );
  319. false ->
  320. timer:sleep(50),
  321. continue
  322. end
  323. end.
  324. flush_inbox() ->
  325. receive
  326. _ -> flush_inbox()
  327. after 0 ->
  328. ok
  329. end.
  330. get_line_from_stacktrace() ->
  331. {current_stacktrace, Stacktrace} = process_info(self(), current_stacktrace),
  332. [{_, _, _, FileInfo} | _] = lists:dropwhile(fun({Module, _Method, _Arity, _FileInfo}) ->
  333. Module =:= ?MODULE end, Stacktrace),
  334. proplists:get_value(line, FileInfo).