Delayed removal from server pool

[William Clements]

* On 04/22/01 list-linux.ha@spoiled.org wrote:
> > Hi,
> > 
> > I've noticed the following with ldirectord (http and https
> > configurations).  ldirectord is setup to use the "negotiate"
> > checktype when querying the real server(s).  I find that
> > if I bring down the real server(s) (or sever their network
> > connection) without first bring down the actual service
> > being monitered (apache in this case), there is a delay of
> > a couple of minutes in ldirectord pulling the affected real 
> > server(s) out of the avaliable server pool.  I've found that 
> > there is still an open connection between the LVS-ldirectord 
> > machine and the real server(s) that were just take off line.  
> > This "connection" is in a SYN_SENT state which apparently takes
> > a couple of minutes to timeout.  During this time, external
> > connections are still being directed by LVS to real server(s)
> > that are no longer avaliable.
> > 
> > Is this expected behavior, if so how can I shorten the timeout
> > while not affectiong anything else?
> 
> Hi William,
> I just had a short look into the man page and maybe "negotiatetimeout" is 
> what you need:
> 
> negotiatetimeout = n
> 
> Defines the number of seconds to wait for TCP/IP timeouts. Default is
> defined by the operating system. If defined in virtual server section
> then the global value is overridden.
> 

I tried this before as well, but noticed no change in the length of the
timeout, but I'll double check (my attention span seems to be much
deminished these days, so I may have missed something).

Much thanks for the info.

Will

> For more information you might also ask on the lvs mailing list, see
> http://www.LinuxVirtualServer.org/
> 
> Cheers,
> Juri
> 
> -- 
> Juri Haberland  <juri@koschikode.com>