Hi We work with the tcp profiles. I dont get your answer 100%.
Our test: unplug the network from one of the backend servers. The light goes red imeadently but all clients connected to this backend server have a timeout of 2 minutes. And it does not help to make a reload of the webbroweser if i use owa. (try to reconnect) Other clients connected to other backend still work… I also set the option to „backend respone time sec = 2“ I dont get why the loadbalancer show the backend as red but needs 2 minutes to redirect client which where connected to this backend. This is not a behavour that should be like this. A friend of mine uses the KEMP loadbalancer and when a backend server dies it switches over in 2-5 seconds… Like this if a server is going down all outlooks connected to the failed backend server will hang for about 2 minutes! Regards! Von: Emilio Campos [mailto:emilio.campos.mar...@gmail.com] Gesendet: Freitag, 12. Juni 2015 13:34 An: zenloadbalancer-support@lists.sourceforge.net Betreff: Re: [Zenloadbalancer-support] backend failover very slow What farm profile are you working with? If TCP: The exchange server will be marked as offline before of the next connection without open, so if the load balancer takes 2 minutes or more to decide to send the next connection to the exchange server downed at the moment zen tries to open a TCP socket and fails is when the server will be marked as down, you are not going to see the backend as down at the same time bakend goes down, only at the time that the load balancer decides to use it and it fails. The maintenance mode disables the backend but it doesn't delete the persistent sessions, so if you are working with persitence mode enabled the next connections are not going to be sent to this backend, but memorized sessions are maintained until they die. Regards! 2015-06-11 17:25 GMT+02:00 VELARTIS Philipp Dürhammer <p.duerham...@velartis.at<mailto:p.duerham...@velartis.at>>: I was testing with our exchange environment and after putting one server offline the zen loadbalancer needs about 2!!! minutes to check that the servers are offline I could verify this with tcp dump. Also i was playing around with maintainance mode. Absolutely nothing happens. I can see in tcp dump that all requests still go to the server in maintainance mode. When i delete a backend interface still connections go the deleted server … only changing the port oft the farm and changing it back helps (seems then pen is restartet) Is this a normal behavior??? ------------------------------------------------------------------------------ _______________________________________________ Zenloadbalancer-support mailing list Zenloadbalancer-support@lists.sourceforge.net<mailto:Zenloadbalancer-support@lists.sourceforge.net> https://lists.sourceforge.net/lists/listinfo/zenloadbalancer-support -- Load balancer distribution - Open Source Project http://www.zenloadbalancer.com Distribution list (subscribe): zenloadbalancer-support@lists.sourceforge.net<mailto:zenloadbalancer-support@lists.sourceforge.net> -- ✓ Geprüft von IKARUS anti.virus http://www.ikarussecurity.com
------------------------------------------------------------------------------
_______________________________________________ Zenloadbalancer-support mailing list Zenloadbalancer-support@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/zenloadbalancer-support
