|
Lanz, Ralf wrote:
>
> Hello everybody,
>
> I am stuck on a problem using lvs, ultramonkey with heartbeat, load
> balancing and Apache (either 1.3.x or 2.x). I did a setup with just two
> servers, where the load balancer (ldirectord) for www resides on the
> same machine as the real servers. The setup that I used is exactly as
> described in http://www.ultramonkey.org/3/topologies/sl-ha-lb-eg.html
> changing actually only the IP-addresses for my setup. I'm using standard
> debian sarge. The setup basically works, the virtual IP is accessable
> and the load is distributed between the two real machines. If any of the
> two machines fail, the other takes the ldiretord.
> The problem is, that whenever apache is running on both machines,
> performance (slightly) degrades, and every other http-request to the
> web-server is lost with timeout (maybe 1 out of four). Deactivating
> either web-server returns the cluster to normal perfomance. I have
> checked all involved logs, even turned on debugging for heartbeat. No
> clarifying messeges or error. The lost packages are lost without any
> log-entry. No firewall is activ. In ipvsadm -L the table seems to be as
> it should:
>
> dvgw-web1:/var/log/apache2# ipvsadm -L
> IP Virtual Server version 1.0.11 (size=4096) Prot LocalAddress:Port
> Scheduler Flags
> -> RemoteAddress:Port Forward Weight ActiveConn InActConn
> TCP dvgw-web:www rr
> -> dvgw-web1:www Local 1 0 0
> -> dvgw-web2:www Route 1 0 3
>
>
> Does anybody have any idea what could be wrong or how to get an
> information on what happens to the lost requests?
It sounds a lot like connections aren't making it to one of the apache
instances. Are you sure that you are handling the arp problem?
--
Horms
H: http://www.vergenet.net/~horms/ W: http://www.valinux.co.jp/en/
|
