|
(Resent, original mail it lost because of sender address check)
Roberto Nibali wrote:
[...]
> You can't :) IP_VS_SO_GET_TIMEOUTS is not implemented in ipvsadm or I'm
> blind, also the proc-fs related entries for this are not exported. I've
> written a patch to re-instate the proper settings in proc-fs, however
> only in 2.4.x kernels. Julian has recently proposed a very granular
> timeout framework, however none of us has had the time nor impulse to
> implement it. For our customers I needed the ability to instrument all
Does that mean we will get timeput parameters _per service_ instead of
global ones? Hooray! When can I test? :)
Roberto Nibali wrote:
[...]
>> Besides that to my knowledge Linux uses a timeout of about 2h + several
>> minutes ( 9 * 75 secs?) until a tcp connection timeouts.
>
>
> Where did you get this number from? Also there is tons of timers
> regarding TCP, so you have to be a bit more specific. On top of that LVS
eh, I got it form /proc:
# cat /proc/sys/net/ipv4/tcp_keepalive_time
7200
And yes, I am talking about Linux clients. I didn't for win yet.
[FYI: http://cryp.to/publications/masquerading-idle-connections/ to see
what I mean with the probes: "... whether a connection needs a
keep-alive packet to be sent.... " ] But it' dependent on the
application, if it usesd the feature, I has to be set when opening a socket.
Best Regards,
Andreas
|
