when i have about 10 users in a muc_room, and about 10 message per second. the ejabberd works fine. but if i login another 100 user with no message send by these 100 man, the problem comes. watch_dog told me:" the process is blocked trying to send data over its TCP connection". then the memory glow up fast. and crashed out of mem.
what can i do for it ?
ps: the mem per user who in a muc_room seem a bit high. for my situation, i need a room serviced for about 10000users and about 1000 message per second.
any infomation for me to deal with?
sorry for my poor english.......
Anyone can help me ?
Anyone can help me ?
c2s or s2s shaper's ? by
c2s or s2s shaper's ?
by default c2s shaper - 1000 bytes per second.
%%% ===============%%% TRAFFIC SHAPERS
%%
%% The "normal" shaper limits traffic speed to 1.000 B/s
%%
{shaper, normal, {maxrate, 1000}}.
%%
%% The "fast" shaper limits traffic speed to 50.000 B/s
%%
{shaper, fast, {maxrate, 50000}}.
%% The "proxy65_shaper" shaper limits traffic speed to 1.000 B/s
{shaper, proxy65_shaper, {maxrate, 1000}}.
i tried 1000000 and unlimit,
i have tried 1000000 and unlimit, but it does not help. the process blocked still.
xmukmb wrote: for my
for my situation, i need a room serviced for about 10000users and about 1000 message per second.
any infomation for me to deal with?
Those numbers seems pretty high, 10000 users, 1000 messages/seconds means ejabberd must have to send 10000000 messages/sec... even assuming that messages are really small, say 56bytes, you still need to send more than 500 MB/s over your network. Are your ejabberd machine and network infrastructure good enough to handle this?
Worse, clustering won't help you much in this situation, as all messages are dispatched by a single MUC room.
i'm sorry, i think i have
i'm sorry, i think i have made some mistake.
the numbers are not so high,there are about 10000users in a room and about 10 message per second.
and for this situation, can someone help me ?