OoO En cette matinée pluvieuse du  mardi 11 octobre 2011, vers 10:39, je
disais:

> I have run some bench with stud + haproxy (1.5-dev7) and found that
> stud with "--write-proxy" parameter adds 40 ms of latency. stunnel has
> recently added (4.45) PROXY support too and enabling it does not add
> any latency. Therefore, I think the problem does not lie in HAProxy
> but maybe there is some bad interaction. I have modified stud to do a
> strcpy of a static proxy line:

>               strcpy(ring_pnt, "PROXY TCP4 1.1.1.1 172.31.10.15 56732
> 443\r\n");
>               written = 43;

> Still, 40 ms of latency. I have also modified stud to not send the
> proxy line until there is something else to send (by commenting out
> "ev_io_start" in "if (OPTIONS.WRITE_PROXY_LINE)" section of
> handle_connect(). No change. Since the code handling PROXY protocol is
> so small in stud, I have no other idea of what to do. See here for the
> relevant code:
>  https://github.com/bumptech/stud/blob/master/stud.c#L489

> Any idea where the latency could lie?

The problem  was stud did not  disable Nagle's algorithm  on the backend
side. Therefore, nothing to do with HAProxy.
-- 
Vincent Bernat ☯ http://vincent.bernat.im

HARDFAIL("Not enough magic.");
        2.4.0-test2 /usr/src/linux/drivers/block/nbd.c

Reply via email to