Linux kernel tuning TCP/Memcached

[theBlueSage]

On Mon, 2010-08-02 at 22:04 -0400, Matthew Gillen wrote:

> On 08/02/2010 07:59 PM, Dan Ritter wrote:
> > http://lartc.org/howto/lartc.kernel.obscure.html
> > 
> > excerpt:
> > 
> >     /proc/sys/net/ipv4/tcp_tw_recycle
> > 
> >     Enable fast recycling TIME-WAIT sockets. Default value is 1.
> >     It should not be changed without advice/request of technical
> >     experts.
> > 
> > Nevertheless, it seems to be set to zero by default on most
> > distros.
> >
> > Hrm. Warrants further investigation.
> 
> Indeed.  A little more digging turned up this:
> 
> http://www.speedguide.net/read_articles.php?id=121
> --------------------------------------
> TCP_TW_RECYCLE
> It enables fast recycling of TIME_WAIT sockets. The default value is 0
> (disabled). The sysctl documentation incorrectly states the default as
> enabled. It can be changed to 1 (enabled) in many cases. Known to cause some
> issues with hoststated (load balancing and fail over) if enabled, should be
> used with caution.
> 
>     echo 1 > /proc/sys/net/ipv4/tcp_tw_recycle
>     (boolean, default: 0)
> --------------------------------------
> 
> It always surprises me how little documentation there is for many kernel
> parameters...
> 


Me too ... I have been at this for a full man-day now and read till my
eyes bled. All that I can find out is that the use of recycling can have
adverse effects on NATing due to the timestamp mechanisms, when using
PAWS .... not very informative. Given all that, I am going to embark on
a huge set of testing, stressing and maxing out hte server and see if
anything breaks. In the end I guess in lieu of experiential wisdom I am
gong to have to go with the trusty old timer : Empirical data :) I will
let the list know how it turns out.  Thanks for the responses

Richard 





> Matt
> _______________________________________________
> Discuss mailing list
> Discuss-mNDKBlG2WHs at public.gmane.org
> http://lists.blu.org/mailman/listinfo/discuss