[lvs-users] Fwd: InActConn dilemma
graeme at graemef.net
Sat Feb 13 23:51:14 GMT 2010
On Sat, 2010-02-13 at 20:48 +0100, Andre Magri wrote:
> Is there a logical reason as to why one of the servers is having so many
> connections being turned into inactive (FIN_WAIT) so quickly?
You are using the WLC scheduler - the server with the least number of
*active* connections gets the next new one.
It looks to me like your two servers are different specs, perform
differently or Apache is configured differently because one of them is
turning connections over far more quickly than the other.
Because you're using -DR, the director never sees the session completely
close so holds sessions in "inactive" state until TCP timeout.
I don't believe you have an LVS problem here; you have a problem that
your two webservers are performing differently. Is on set to use Apache
KeepAlives, by any chance?
More information about the lvs-users