|
On Wed, Sep 21, 2005 at 08:04:41PM +0530, Vijay K wrote:
> Hi
> I have set up two directord with two ethernet interfaces each..each in a
> different network.
>
> Here are the details:
> director1 : 10.10.30.35 and 10.10.33.1
> dir2 : 10.10.30.36 and 10.10.33.2
> virtual : 10.10.30.30 and 10.10.33.33
>
> the real servers are in the 33 network.. they r 10.10.33.11, 10.10.33.12,
> 10.10.33.13..etc
>
> Now the prob i have is that even though its ucast and ip of teh other
> machien is given i end up having udp packets on both teh networks.Arents
> they supposed to go only to teh other director?
>
> My ha.cf file of director2 looks like this:(the one for director 1 looks
> similar except for the corresponding ip address changes)
>
> debugfile /var/log/ha-debug
> logfile /var/log/ha-log
> logfacility local0
> keepalive 2
> deadtime 10
> baud 19200
> udpport 694
> udp eth0 eth1
> auto_failback off
> ucast eth0 10.10.30.35
> ucast eth1 10.10.33.1
> node dir2.xxxx.com
> node dir1.xxxx.com
>
> udp for eth0 and eth1 specified.
> So can somebody tell me why the udp packets are going to all teh other
> machines in bothe 10.10.30. and 10.10.33. networks. Is this teh norm or shud
> i do something to avoid the udp packets bothering other machiens in the
> network.
Your configuration probably sets up heartbeat to use broadcast
on eth0 and eth1, which is why it is doing just that. If you
are concerned about disturbing other machines on the network,
which is a valid concern IMHO, then consider using mcast (multicast) or
ucast (unicast) instead of bcast (broadcast) in your configuation.
> Also,the health check from directord to realservers( ARP) are going to the
> 10.10.30 network(for those realservers mentioned in configuration file and
> not present in the network).but these arp requests are going to teh machiens
> in 10.10.30 network also checking/requesting if they r the realservers
I'm not entirely sure what health checks you are talking about here,
but if you are talking about checks generated by ldirectord, keepalived
or something like that, directed to real-servers of an LVS
virtual-server, then these are typically sent using TCP, and thus the
routing rules on the machine that is running the tests will dictate
which interface is used.
--
Horms
|
