diff options
author | Anders Svensson <[email protected]> | 2013-02-27 19:28:38 +0100 |
---|---|---|
committer | Anders Svensson <[email protected]> | 2013-03-04 17:41:37 +0100 |
commit | 21708fe99473c002cf2f89c1761ba90485bb802d (patch) | |
tree | c7154557b3a33b1abf0913d288c5a404426aa88a /lib | |
parent | 3e77e534655b7197a0999a68480fb524b8cb0273 (diff) | |
download | otp-21708fe99473c002cf2f89c1761ba90485bb802d.tar.gz otp-21708fe99473c002cf2f89c1761ba90485bb802d.tar.bz2 otp-21708fe99473c002cf2f89c1761ba90485bb802d.zip |
Rework watchdog suite to remove most timing dependencies
Count watchdog messages instead when this is an alternative.
Diffstat (limited to 'lib')
-rw-r--r-- | lib/diameter/test/diameter_watchdog_SUITE.erl | 270 |
1 files changed, 143 insertions, 127 deletions
diff --git a/lib/diameter/test/diameter_watchdog_SUITE.erl b/lib/diameter/test/diameter_watchdog_SUITE.erl index e1e166b834..388177a0bf 100644 --- a/lib/diameter/test/diameter_watchdog_SUITE.erl +++ b/lib/diameter/test/diameter_watchdog_SUITE.erl @@ -30,7 +30,7 @@ end_per_suite/1]). %% testcases --export([reopen/1, reopen/4, reopen/7]). +-export([reopen/1, reopen/4, reopen/6]). -export([id/1, %% jitter callback run1/1]). @@ -64,7 +64,7 @@ {'Host-IP-Address', [?ADDR]}, {'Vendor-Id', 42}, {'Product-Name', "OTP/diameter"}, - {'Auth-Application-Id', [?DIAMETER_APP_ID_COMMON]}, + {'Auth-Application-Id', [0 = ?BASE:id()]}, {application, [{alias, Name}, {dictionary, ?BASE}, {module, ?MODULE}]}]). @@ -72,45 +72,38 @@ %% Watchdog timer as a callback. -define(WD(T), {?MODULE, id, [T]}). -%% Watchdog timers used by the testcases. Note that the short timeout -%% with random jitter is excluded since the reopen/1 isn't smart -%% enough to deal with it: see ONE_WD below. --define(WD_TIMERS, [?WD(6000) - | [F_(T_) || T_ <- [10000, 20000, 30000], - F_ <- [fun(T__) -> T__ end, - fun(T__) -> ?WD(T__) end]]]). - -%% Watchdog timer of the misbehaving peer. --define(PEER_WD, 10000). - -%% Receive a watchdog event within a specified time. --define(EVENT(T, Tmo), - receive #diameter_event{info = T} -> now() - after Tmo -> ?ERROR({timeout, Tmo}) - end). - -%% Receive an event in a given number of watchdogs, plus or minus -%% half. Note that the call to now_diff assumes left to right -%% evaluation order. --define(EVENT(T, N, WdL, WdH), - [?ERROR({received, _Elapsed_, _LowerBound_, N, WdL}) - || _UpperBound_ <- [(N)*(WdH) + (WdH) div 2], - _Elapsed_ <- [now_diff(now(), ?EVENT(T, _UpperBound_))], - _LowerBound_ <- [(N)*(WdL) - (WdL) div 2], - _Elapsed_ =< _LowerBound_*1000]). - --define(EVENT(T, N, Wd), - ?EVENT(T, N, Wd, Wd)). - -%% A timeout that ensures one watchdog. The ensure only one watchdog +%% Watchdog timers used by the testcases. +-define(WD_TIMERS, [10000, ?WD(10000)]). + +%% Watchdog timer of the misbehaving node. Longer than the peer's so +%% that DWR is never sent at watchdog timeout. +-define(PEER_WD, 20000). + +%% A timeout that ensures one watchdog. To ensure only one watchdog %% requires (Wd + 2000) + 1000 < 2*(Wd - 2000) ==> 7000 < Wd for the %% case with random jitter. -define(ONE_WD(Wd), jitter(Wd,2000) + 1000). +%% Receive an event message from diameter. +-define(EVENT(T), + apply(fun() -> %% apply to not bind T_ + receive #diameter_event{info = T = T_} -> + log_event(T_) + end + end, + [])). + +%% Receive a watchdog event. +-define(WD_EVENT(Ref), log_wd(element(4, ?EVENT({watchdog, Ref, _, _, _})))). + +%% Log to make failures identifiable. +-define(LOG(T), ?LOG("~p", [T])). +-define(LOG(F,A), ct:pal("~p: " ++ F, [self() | A])). + %% =========================================================================== suite() -> - [{timetrap, {minutes, 10}}].%% enough for 17 watchdogs @ 30 sec plus jitter + [{timetrap, {minutes, 5}}]. %% 20 watchdogs @ 15 sec all() -> [reopen]. @@ -130,82 +123,67 @@ end_per_suite(_Config) -> %% and reopen behaviour by examining watchdog events. reopen(_) -> - [] = run([[reopen, T, Wd, N, M] - || Wd <- ?WD_TIMERS, %% watchdog_timer value - T <- [listen, connect], %% watchdog to test + [] = run([[reopen, T, W, N, M] + || T <- [listen, connect], %% watchdog to test + W <- ?WD_TIMERS, %% watchdog_timer value N <- [0,1,2], %% DWR's to answer before ignoring M <- ['DWR', 'DWA', 'RAA']]). %% how to induce failback -reopen(Type, Wd, N, M) -> - Server = start_service(), - Client = start_service(), - - %% The peer to the transport whose watchdog is tested is given a - %% long watchdog timeout so that it doesn't send DWR of its own. - {Node, Peer} = {{[], Wd}, {[{module, ?MODULE}], ?WD(?PEER_WD)}}, - - {{LH,LW},{CH,CW}} = case Type of - listen -> {Node, Peer}; - connect -> {Peer, Node} - end, - - LO = [{transport_module, diameter_tcp}, - {transport_config, LH ++ [{ip, ?ADDR}, {port, 0}]}, - {watchdog_timer, LW}], - - {ok, LRef} = diameter:add_transport(Server, {listen, LO}), +reopen(Test, Wd, N, M) -> + %% Publish a ref ensure the connecting transport is added only + %% once events from the listening transport are subscribed to. + Ref = make_ref(), + [] = run([[reopen, T, Test, Ref, Wd, N, M] || T <- [listen, connect]]). - [LP] = ?util:lport(tcp, LRef, 20), - - CO = [{transport_module, diameter_tcp}, - {transport_config, CH ++ [{ip, ?ADDR}, {port, 0}, - {raddr, ?ADDR}, {rport, LP}]}, - {watchdog_timer, CW}], - - %% Use a temporary process to ensure the connecting transport is - %% added only once events from the listening transport are - %% subscribed to. - Pid = spawn(fun() -> receive _ -> ok end end), +%% reopen/6 - [] = run([[reopen, Type, T, LRef, Pid, Wd, N, M] - || T <- [{listen, Server}, {connect, Client, CO}]]). +reopen(Type, Test, Ref, Wd, N, M) -> + {SvcName, TRef} = start(Type, Ref, opts(Type, Test, Ref, Wd)), + reopen(Type, Test, SvcName, TRef, Wd, N, M). -%% start_service/1 +%% start/3 -start_service() -> +start(Type, Ref, Opts) -> Name = hostname(), + true = diameter:subscribe(Name), ok = diameter:start_service(Name, [{monitor, self()} | ?SERVICE(Name)]), - Name. - -%% reopen/7 - -reopen(Type, {listen = T, SvcName}, Ref, Pid, Wd, N, M) -> - diameter:subscribe(SvcName), - Pid ! ok, - recv(Type, T, SvcName, Ref, Wd, N, M); + {ok, TRef} = diameter:add_transport(Name, {Type, Opts}), + true = diameter_reg:add_new({Type, Ref, Name}), + {Name, TRef}. + +opts(Type, Test, Ref, Wd) -> + {Timer, Counts, Mod} = opts(Type, Test, Wd), + [{transport_module, diameter_tcp}, + {transport_config, Mod ++ [{ip, ?ADDR}, {port, 0}] ++ cfg(Type, Ref)}, + {watchdog_timer, Timer}, + {watchdog_counts, Counts}]. + +cfg(listen, _) -> + []; +cfg(connect, Ref) -> + [{{_, _, SvcName}, _Pid}] = diameter_reg:wait({listen, Ref, '_'}), + [[{ref, LRef} | _]] = diameter:service_info(SvcName, transport), + [LP] = ?util:lport(tcp, LRef, 20), + [{raddr, ?ADDR}, {rport, LP}]. -reopen(Type, {connect = T, SvcName, Opts}, _, Pid, Wd, N, M) -> - diameter:subscribe(SvcName), - MRef = erlang:monitor(process, Pid), - receive {'DOWN', MRef, process, _, _} -> ok end, - {ok, Ref} = diameter:add_transport(SvcName, {T, Opts}), - recv(Type, T, SvcName, Ref, Wd, N, M). +opts(Type, Type, Wd) -> + {Wd, [], []}; +opts(_Type, _Test, _Wd) -> + {?WD(?PEER_WD), [{open, 0}], [{module, ?MODULE}]}. -%% recv/7 +%% reopen/7 %% The watchdog to be tested. -recv(Type, Type, _SvcName, Ref, Wd, N, M) -> +reopen(Type, Type, SvcName, Ref, Wd, N, M) -> + ?LOG("node, ~p, ~p", [Type, Ref]), + %% Connection should come up immediately as a consequence of %% starting the watchdog process. In the accepting case this %% results in a new watchdog on a transport waiting for a new %% connection. - ?EVENT({watchdog, Ref, _, {initial, okay}, _}, 2000), - ?EVENT({up, Ref, _, _, #diameter_packet{}}, 0), - - %% Low/high watchdog timeouts. - WdL = jitter(Wd, -2000), - WdH = jitter(Wd, 2000), + {initial, okay} = ?WD_EVENT(Ref), + ?EVENT({up, Ref, _, _, #diameter_packet{}}), %% OKAY Timer expires & Failover() %% Pending SetWatchdog() SUSPECT @@ -215,8 +193,13 @@ recv(Type, Type, _SvcName, Ref, Wd, N, M) -> %% the first unanswered DWR. Knowing the min/max watchdog timeout %% values gives the time interval in which the event is expected. - ?EVENT({watchdog, Ref, _, {okay, suspect}, _}, N+2, WdL, WdH), - ?EVENT({down, Ref, _, _}, 0), + [0,0,0,0] = wd_counts(SvcName), + + {okay, suspect} = ?WD_EVENT(Ref), + ?EVENT({down, Ref, _, _}), + + %% N received DWA's + [_,_,_,N] = wd_counts(SvcName), %% SUSPECT Receive DWA Pending = FALSE %% Failback() @@ -228,8 +211,13 @@ recv(Type, Type, _SvcName, Ref, Wd, N, M) -> %% The peer sends a message before the expiry of another watchdog %% to induce failback. - ?EVENT({watchdog, Ref, _, {suspect, okay}, _}, WdH + 2000), - ?EVENT({up, Ref, _, _}, 0), + {suspect, okay} = ?WD_EVENT(Ref), + ?EVENT({up, Ref, _, _}), + + %% N+1 sent DWR's, N/N+1 received DWA's + R1 = N+1, + A1 = choose(M == 'DWA', R1, N), + [R1,_,_,A1] = wd_counts(SvcName), %% OKAY Timer expires & SendWatchdog() %% !Pending SetWatchdog() @@ -242,16 +230,19 @@ recv(Type, Type, _SvcName, Ref, Wd, N, M) -> %% back down after either one or two watchdog expiries, depending %% on whether or not DWA restored the connection. - F = choose(M == 'DWA', 2, 1), - ?EVENT({watchdog, Ref, _, {okay, suspect}, _}, F, WdL, WdH), - ?EVENT({down, Ref, _, _}, 0), + {okay, suspect} = ?WD_EVENT(Ref), + ?EVENT({down, Ref, _, _}), %% SUSPECT Timer expires CloseConnection() %% SetWatchdog() DOWN %% %% Non-response brings the connection down after another timeout. - ?EVENT({watchdog, Ref, _, {suspect, down}, _}, 1, WdL, WdH), + {suspect, down} = ?WD_EVENT(Ref), + + R2 = R1 + choose(M == 'DWA', 1, 0), + A2 = A1, + [R2,_,_,A2] = wd_counts(SvcName), %% DOWN Timer expires AttemptOpen() %% SetWatchdog() DOWN @@ -263,7 +254,7 @@ recv(Type, Type, _SvcName, Ref, Wd, N, M) -> %% %% The connection is reestablished after another timeout. - recv_reopen(Type, Ref, WdL, WdH), + recv_reopen(Type, Ref), %% REOPEN Receive non-DWA Throwaway() REOPEN %% @@ -281,18 +272,27 @@ recv(Type, Type, _SvcName, Ref, Wd, N, M) -> %% An exchange of 3 watchdogs (the first directly after %% capabilities exchange) brings the connection back up. - ?EVENT({watchdog, Ref, _, {reopen, okay}, _}, 2, WdL, WdH), - ?EVENT({up, Ref, _, _, #diameter_packet{}}, 0), + {reopen, okay} = ?WD_EVENT(Ref), + ?EVENT({up, Ref, _, _, #diameter_packet{}}), + + %% Three DWR's have been answered. + R3 = R2 + 3, + A3 = A2 + 3, + [R3,_,_,A3] = wd_counts(SvcName), %% Non-response brings it down again. - ?EVENT({watchdog, Ref, _, {okay, suspect}, _}, 2, WdL, WdH), - ?EVENT({down, Ref, _, _}, 0), - ?EVENT({watchdog, Ref, _, {suspect, down}, _}, 1, WdL, WdH), + {okay, suspect} = ?WD_EVENT(Ref), + ?EVENT({down, Ref, _, _}), + {suspect, down} = ?WD_EVENT(Ref), + + R4 = R3 + 1, + A4 = A3, + [R4,_,_,A4] = wd_counts(SvcName), %% Reestablish after another watchdog. - recv_reopen(Type, Ref, WdL, WdH), + recv_reopen(Type, Ref), %% REOPEN Timer expires & NumDWA = -1 %% Pending & SetWatchdog() @@ -305,36 +305,57 @@ recv(Type, Type, _SvcName, Ref, Wd, N, M) -> %% Peer is now ignoring all watchdogs go down again after 2 %% timeouts. - ?EVENT({watchdog, Ref, _, {reopen, down}, _}, 2, WdL, WdH); + {reopen, down} = ?WD_EVENT(Ref); %% The misbehaving peer. -recv(_, Type, SvcName, Ref, Wd, N, M) -> +reopen(Type, _, SvcName, Ref, Wd, N, M) -> + ?LOG("peer, ~p, ~p", [Type, Ref]), + %% First transport process. - ?EVENT({watchdog, Ref, _, {initial, okay}, _}, 1000), - ?EVENT({up, Ref, _, _, #diameter_packet{}}, 0), + {initial, okay} = ?WD_EVENT(Ref), + ?EVENT({up, Ref, _, _, #diameter_packet{}}), + reg(Type, Ref, SvcName, {SvcName, {Wd,N,M}}), - ?EVENT({watchdog, Ref, _, {okay, down}, _}, infinity), + + {okay, down} = ?WD_EVENT(Ref), %% Second transport process. - ?EVENT({watchdog, Ref, _, {_, reopen}, _}, infinity), + ?EVENT({watchdog, Ref, _, {_, reopen}, _}), reg(Type, Ref, SvcName, 3), - ?EVENT({watchdog, Ref, _, {_, down}, _}, infinity), + ?EVENT({watchdog, Ref, _, {_, down}, _}), %% Third transport process. - ?EVENT({watchdog, Ref, _, {_, reopen}, _}, infinity), + ?EVENT({watchdog, Ref, _, {_, reopen}, _}), reg(Type, Ref, SvcName, 0), - ?EVENT({watchdog, Ref, _, {_, down}, _}, infinity), + ?EVENT({watchdog, Ref, _, {_, down}, _}), ok. -%% recv_reopen/4 +log_wd({From, To} = T) -> + ?LOG("~p -> ~p", [From, To]), + T. + +log_event(E) -> + T = element(1,E), + T == watchdog orelse ?LOG("~p", [T]), + E. + +%% wd_counts/1 -recv_reopen(connect, Ref, WdL, WdH) -> - ?EVENT({watchdog, Ref, _, {_, reopen}, _}, 1, WdL, WdH), - ?EVENT({reconnect, Ref, _}, 0); +wd_counts(SvcName) -> + [Info] = diameter:service_info(SvcName, transport), + {_, Counters} = lists:keyfind(statistics, 1, Info), + [proplists:get_value({{0,280,R}, D}, Counters, 0) || D <- [send,recv], + R <- [1,0]]. -recv_reopen(listen, Ref, _, _) -> - ?EVENT({watchdog, Ref, _, {_, reopen}, _}, 1, ?PEER_WD). +%% recv_reopen/2 + +recv_reopen(connect, Ref) -> + {down, reopen} = ?WD_EVENT(Ref), + ?EVENT({reconnect, Ref, _}); + +recv_reopen(listen, Ref) -> + {_, reopen} = ?WD_EVENT(Ref). %% reg/4 %% @@ -509,11 +530,6 @@ run1([F|A]) -> S end. -%% now_diff/2 - -now_diff(T1, T2) -> - timer:now_diff(T2, T1). - %% jitter/2 jitter(?WD(T), _) -> |