|
>I may have not explained myself correctly and I'm sorry for that :
>The issue is not with ldirectors failover, but rather different behavior of
>connections on real server that ldirector master is running on, and the
real server that does not run ldirector (or is being ldirector stand by).
>On the real server that also has ldirector running, the connections are
kept
>alive although TCP timeout has passed, which is good for my application
which looks the specific socket (ipvsadm -Lc does not show them though only
lsof -i).
>On the other server, that does not run ldirector (or ldirector is in stand
by mode), the application on the client side hangs after TCP timeout
,because the ldirector >will create a new connection (new socket) for the
request that comes from the client.
>I'm using ldirectors with master and backup daemons up , but it has no
affect of the behavior I have described.
>My question is : Is there away to replicate the (although strange)
behavior of 'supposed to be dead but are alive' connections on the other
real server as well?
It sounds like you want persistent client connections. Why not just bump up
the "persistent=" value to however long your application holds the socket
open on the real server, or did I miss something?
I Don't know why it is different between the Director and the RS though...
--Karl
|
