论坛首页 综合技术论坛

hibernate your gen_fsm process

浏览 3369 次
精华帖 (0) :: 良好帖 (0) :: 新手帖 (0) :: 隐藏帖 (0)
作者 正文
   发表时间:2009-12-14  
关于erlang进程hibernate的好处,去过这次erlang大会的同学应该已经都知道了

简单的理解:
hibernate可以使你闲置的erlang进程马上进行gc,因为进程处于receive状态下是不会gc的
我在性能调优时发现,gc对于消息发送速度的影响还是非常大的

关于erlang gc问题也可以参看这个
引用

Recently, as part of RabbitMQ server development, we ran into an interesting issue regarding Erlang’s per-process garbage collection. If a process is idle — not doing any work at all, simply waiting for an external event — then its garbage-collector will not run until it starts working again. The solution is to hibernate idle processes, which causes a very aggressive garbage-collection run and puts the process into a suspended state, from which it will wake when a message next arrives.


所以查看rabbitMQ源代码可以看到里面的gen_server/fsm等进程统统都会使用hibernate

下面解释下如何让进程hibernate,这里需要注意的是一个进程hiberate后,会清空stack栈,也就是说之前的调用关系全部没有了,也就是说hibernate所在的函数永远不会返回,待有新消息时,进程从hibernate(M,F,A)里指定的M:F处开始运行

gen_server和gen_fsm都提供了hibernate的方式,2种方法

第一种:
在gen_server,或gen_fsm回调接口返回时,指定hibernate
{next_state, NStateName, NStateData, Time1} 

实际这个Time1可以指定为hibernate,则重新回到main loop时会直接hibernate

这种方法可能使你的进程频繁hibernate又被唤醒,效率不是很好

正确的方式应该是,当你预期你的进程一段时间内不会收到消息才hibernate

所以推荐下面的方法,直接看下示例代码:
%% ====================================================================
%% Interface Function
%% ====================================================================
join(UserId)->
	io:format("join UserId=~p~n",[UserId]),
	gen_fsm:sync_send_event(?MODULE, {join,UserId}).

start()->
	gen_fsm:send_event(?MODULE,start).

%% ====================================================================
%% Callback Function
%% ====================================================================
wait_player({join,UserId},_From, State) ->
	print_inner_data(State),
	put(p,get(p)+1),
	ets:insert(State#state.players,{p,UserId}),
	Count=State#state.count+1,
	if
		Count =:= 3 ->
			io:format("jump to next phrase -> start~n"),
			{reply,next_phrase,wait_start,State,5000};
		true->
			NewState=State#state{count=Count},
			{reply,stand_still,wait_player,NewState,5000}
	end.
wait_player(timeout,State)->
	io:format("timeout happened when wait_player,let's hibernate~n"),
	proc_lib:hibernate(gen_fsm, enter_loop, [?MODULE, [], wait_player,State]).

wait_start(start,State)->
	 print_inner_data(State),
	 io:format("recv start event ~n"),
	 {next_state,wait_start,State,5000};

wait_start(timeout,State)->
	io:format("timeout happened when wait_start,let's hibernate~n"),
	proc_lib:hibernate(gen_fsm, enter_loop, [?MODULE, [], wait_start,State]).

start_link()->
	gen_fsm:start_link({local,?MODULE}, ?MODULE, [] ,[]).

init([])->
	put(p,0),
	{ok, 
		wait_player, 
		#state
		{
			count=0,
			players=ets:new(players,[bag])
		},
		5000
	}.

print_inner_data(State)->
	io:format("inner data: dict=~p,ets=~p~n",[get(p),ets:tab2list(State#state.players)]).



运行:
2> {ok,Pid}=fsm:start_link().
{ok,<0.39.0>}

3> process_info(Pid,total_heap_size).
{total_heap_size,233}

4> fsm:join(123).
join UserId=123
inner data: dict=0,ets=[]
stand_still
timeout happened when wait_player,let's hibernate

5> process_info(Pid,total_heap_size).
{total_heap_size,33}

6> fsm:join(456).
join UserId=456
inner data: dict=1,ets=[{p,123}]
stand_still


总体上是让进程在main_loop里先timeout也就是证明一段时间内都没有消息到达,然后gen_fsm进程是发送{'gen_event',timeout}消息处理timeout事件的,所以要在每个Mod:StateName处都加个处理timeout的分支,来进行timeout处理,也就是hibernate掉自己,重新进入的是enter_loop 函数,这个东西很有用呐,大家可以查看源码,我就不多说了

另外,示例中可以看到,hibernate会清空stack,但是不会影响你的进程字典和ets等其它东西,可以放心用了










   发表时间:2009-12-18  
gen_*系列 都支持hibernate, 会节约很多内存的...
0 请登录后投票
   发表时间:2010-01-14  
调用栈会清空,但是函数参数就是状态还是在的哦,good job
0 请登录后投票
论坛首页 综合技术版

跳转论坛:
Global site tag (gtag.js) - Google Analytics