diff --git a/src/db_daterange.erl b/src/db_daterange.erl
index 4774487..897ddac 100644
--- a/src/db_daterange.erl
+++ b/src/db_daterange.erl
@@ -42,18 +42,19 @@ insert(Conn, Hash) when is_list(Hash) ->
 	gen_server:cast(srv_name(), insert),
 	IRet.
 
-% [ListHash, ListHash]
+% [{ListHash, Req}, {ListHash, Req}]
 lookup(Conn, DaySecs, Count) ->
 	Sel = {'$query', {date, DaySecs}, '$orderby', {reqs, -1}},
 	List = mongo:do(safe, master, Conn, ?DBNAME, fun() ->
-		Cursor = mongo:find(?COLLNAME, Sel, {'_id', 1}, 0, Count), 
+		Cursor = mongo:find(?COLLNAME, Sel, {'_id', 1, reqs, 1}, 0, Count), 
 		mongo_cursor:rest(Cursor)
 	end),
 	[decode_hash(Doc) || Doc <- List].
 
 decode_hash(Doc) ->
 	{ID} = bson:lookup('_id', Doc),
-	binary_to_list(ID).
+	{Req} = bson:lookup(reqs, Doc),
+	{binary_to_list(ID), Req}.
 		
 % delete all oldest hashes 
 try_delete_oldest(Conn) ->
diff --git a/src/db_store_mongo.erl b/src/db_store_mongo.erl
index 7b5c979..2bfc699 100644
--- a/src/db_store_mongo.erl
+++ b/src/db_store_mongo.erl
@@ -91,9 +91,16 @@ search_newest_top(Conn, Count, DaySecs) ->
 % use hash_date to search, based on date index, so the search speed is very fast
 search_newest_top_by_date(Conn, Count, DaySecs) ->
 	Hashes = db_daterange:lookup(Conn, DaySecs, Count),
-	Infos = [index(Conn, Hash) || Hash <- Hashes],
+	Infos = [day_info(Conn, Hash, Req) || {Hash, Req} <- Hashes],
 	lists:filter(fun(E) -> E /= {} end, Infos).
 
+day_info(Conn, Hash, Req) ->
+	case index(Conn, Hash) of
+		{} -> {};
+		Info ->
+			setelement(4, Info, Req)
+	end.
+
 index(Conn, Hash) when is_list(Hash) ->
 	Ret = mongo_do_slave(Conn, fun() ->
 		mongo:find_one(?COLLNAME, {'_id', list_to_binary(Hash)})
@@ -214,6 +221,8 @@ decode_ret_item(Item) ->
 	{Torrent} = bson:lookup(obj, Item),
 	decode_torrent_item(Torrent).
 
+% {single, Hash, {Name, Length}, RequestCount, CreatedAt}
+% {multi, Hash, {Name, Files}, RequestCount, CreatedAt}
 decode_torrent_item(Torrent) ->	
 	{BinHash} = bson:lookup('_id', Torrent),
 	Hash = binary_to_list(BinHash),
diff --git a/src/torrent/tor_download.erl b/src/torrent/tor_download.erl
index b44ca1a..f36a8aa 100644
--- a/src/torrent/tor_download.erl
+++ b/src/torrent/tor_download.erl
@@ -23,15 +23,15 @@
 -define(REQ_TIMEOUT, 60*1000).
 % when ibrowse crashed, it will not notify these requests timeout, that will
 % make these requests stay in the state forever
--define(REQ_ERROR_TIMEOUT, 2*?REQ_TIMEOUT).
+-define(REQ_ERROR_TIMEOUT, 5*60*1000).
 -define(IS_ERROR_TIMEOUT(T), (timer:now_diff(now(), T) div 1000 > ?REQ_ERROR_TIMEOUT)).
 -record(state, {start, hashSum = 0, reqSum = 0, totalTime = 0, reqs}).
 
 start_global() ->
 	ibrowse:start(),
-	Options = [{max_sessions, ?HTTP_SESSION}, {max_pipeline_size, ?HTTP_PIPELINE}],
+	%Options = [{max_sessions, ?HTTP_SESSION}, {max_pipeline_size, ?HTTP_PIPELINE}],
 	% not work here ?
-	[ibrowse:set_dest(Host, 80, Options) || Host <- get_req_hosts()],
+	%[ibrowse:set_dest(Host, 80, Options) || Host <- get_req_hosts()],
 	ok.
 
 start_link() ->
@@ -48,14 +48,14 @@ stats(Pid) ->
 	gen_server:call(Pid, get_stats, infinity).
 
 init([]) ->
-	{ok, #state{start = now(), reqs = gb_trees:empty()}}.
+	{ok, #state{start = now(), reqs = gb_trees:empty()}, 0}.
 
 handle_cast({download, MagHash, From}, State) ->
 	#state{reqs = Reqs, hashSum = H, reqSum = R} = State,
 	% remove these invalid requests
 	UpdateReqs = Reqs, %check_error_timeout_reqs(Reqs),
 	NewReqs = create_download(UpdateReqs, MagHash, From),
-	NewSum = R + 1 - gb_trees:size(Reqs) - gb_trees:size(UpdateReqs),
+	NewSum = R + 1 - (gb_trees:size(Reqs) - gb_trees:size(UpdateReqs)),
 	{noreply, State#state{reqs = NewReqs, hashSum = H + 1, reqSum = NewSum}};
 
 handle_cast(stop, State) ->
@@ -93,6 +93,10 @@ handle_info({ibrowse_async_response, ReqID, Body}, State) ->
 	end,
 	{noreply, NewState};
 
+handle_info(timeout, State) ->
+	timer:send_interval(?REQ_ERROR_TIMEOUT div 2, check_error_timeout),
+	{noreply, State};
+
 handle_info(check_error_timeout, State) ->
 	#state{reqs = Reqs} = State,
 	NewReqs = check_error_timeout_reqs(Reqs),
@@ -164,19 +168,20 @@ unzip_content(_B) ->
 	error.
 
 %% http stuff
-get_req_hosts() ->
-	["http://bt.box.n0808.com", 
-   	"http://torcache.net",
-   	"http:/torrange.com",
-   	"http://zoink.it"].
+%get_req_hosts() ->
+%	["http://bt.box.n0808.com", 
+%   	"http://torcache.net",
+%   	"http:/torrange.com",
+%   	"http://zoink.it"].
 
 create_req_urls(MagHash) when is_list(MagHash), length(MagHash) == 40 ->
 	U1 = "http://torcache.net/torrent/" ++ MagHash ++ ".torrent",
 	U2 = format_btbox_url(MagHash),
 	U3 = "http://torrage.com/torrent/" ++ MagHash ++ ".torrent",
+	% zoink cause ibrowse crash because error response
 	% zoink.it support https, but the ssl library seems memory leak
-	U4 = "http://zoink.it/torrent/" ++ MagHash ++ ".torrent",
-	[U1, U2, U3, U4].	
+	%U4 = "http://zoink.it/torrent/" ++ MagHash ++ ".torrent",
+	[U1, U2, U3].	
 
 is_ssl_url(URL) when is_list(URL), length(URL) > 4 ->
 	string:substr(URL, 1, 5) == "https".
@@ -187,16 +192,26 @@ format_btbox_url(MagHash) ->
 	"http://bt.box.n0808.com/" ++ H ++ "/" ++ T ++ "/" ++ MagHash ++ ".torrent".
 
 check_error_timeout_reqs(Reqs) ->
+	Size = gb_trees:size(Reqs),
 	ReqList = gb_trees:to_list(Reqs),
-	lists:foldl(fun(E, NewReqs) ->
+	NewReqs = lists:foldl(fun(E, NewReqs) ->
 		check_error_timeout(NewReqs, E)
-	end, gb_trees:empty(), ReqList).
+	end, gb_trees:empty(), ReqList),
+	NewSize = gb_trees:size(NewReqs),
+	case NewSize < Size of
+		true ->
+			?E(?FMT("remove ~p timeout download reqs, new size ~p", 
+				[Size - NewSize, NewSize]));
+		false ->
+			ok
+	end,
+	NewReqs.
 
 check_error_timeout(Acc, {ReqID, {MagHash, _, From, _, Start} = Req}) ->
 	case ?IS_ERROR_TIMEOUT(Start) of
 		true ->
 			From ! {got_torrent, failed, MagHash},
-			?E(?FMT("download req error timeout ~s", [MagHash])),
+			?E(?FMT("download req timeout ~p ~s", [ReqID, MagHash])),
 			Acc;
 		false ->
 			gb_trees:insert(ReqID, Req, Acc)