Hi Henry,

We run our servers and the larger gateways all on one vlan and phones
are on other vlans based roughly on their native departmental subnets.
 Cisco does per-vlan spanning tree which is handy in the case that
someone loops a switch, they only take down that vlan.

For Cisco we use the "ip helper-address" command under the vlan
interfaces.  I point the helper address at my master server, and use
the built-in DHCP fail-over described here:
DHCP is broadcast so any DHCP server on the subnet will respond, ie
your secondary server (Be careful, if you have a test server it could
also be giving out addresses).

http://barryp.org/blog/entries/dhcp-failover/

Here is an example of a subnet config under the dhcp.master file:

# Global Options
option domain-name "example.com";
option domain-name-servers 10.100.0.10,10.100.0.11;
option tftp-server-name "10.100.0.10";
option time-servers 10.100.0.10,10.100.0.11;
option ntp-servers 10.100.0.10,10.100.0.11;
option time-offset -25200;
default-lease-time 86400;
max-lease-time 86400;
authoritative;
ddns-update-style  none;
option sip-servers-name code 120 = text;

subnet 10.100.106.0 netmask 255.255.255.0 {
        pool {
                failover peer "dhcp";
                deny dynamic bootp clients;
                range 10.100.106.12 10.100.106.254;
                option subnet-mask 255.255.255.0;
                option broadcast-address 10.100.106.255;
                option routers 10.100.106.1;
                default-lease-time 86400;
                max-lease-time 86400;
                        group {
                        next-server sipx1.example.com;
# TFTP server
                                host engines_lab_fax_fxs {
                                        hardware ethernet 00:90:8f:00:00:00;
                                        fixed-address 10.100.106.10;
                                }
                        }
        }
}

I use rsync to keep them sync'd as the dhcpd.master file is the same
on both servers (dhcpd.conf varies)

As Matt said DNS is critical, I run it on both servers and use a DNS
forward for things that don't resolve locally (so you can do yum
updates), also I run NTP on both servers (they in turn sync
externally)  my philosophy is to keep as much critical traffic to VoIP
internal to the two servers as much as possible.  Also a tip about
DNS, the default SIPX HA install sets up the slave server to use the
DNS slaving feature.  I don't like this for my setup as I've found
that unless you're very careful on how your set it up the DNS slave
can stop responding to requests if it hasn't gotten an update from
some master server after a configurable time-out period.  So I run
both the master and the slave sipx server in master DNS mode and keep
the zones sync'd.


Regards,

Kyle


On Tue, Jan 4, 2011 at 6:38 AM, Matt White <[email protected]> wrote:
>>>> "Henry Dogger" <[email protected]> 01/04/11 3:45 AM >>>
>>>>Hi Kyle,
>>>>
>>>>Thanks for the quick reply :)
>>>>
>>>>So you have DHCP and DNS on both sipx boxes? It would seem to me you
>>>>only need one DHCP right?
>>>>My DNS, DHCP, NTP and TFTP are all on the master server.
>
> If DNS is only on your master node, and the master node is down then calls 
> will fail.  DNS is required.  DHCP on just the master works as long as the 
> phones DHCP lease hasn't expired or isn't rebooted.  Once the lease is lost, 
> it will never get a new ip.
>
> DNS is critical.  Each endpoint should should be configured for 2 dns 
> servers.  And those DNS servers must have SRV records for both the master 
> node and the slave node.
>
> Even if you have two dns servers available, but the SRV records for SIP only 
> reference the master node, then calls will still fail.
>
> -M
>
> _______________________________________________
> sipx-users mailing list
> [email protected]
> List Archive: http://list.sipfoundry.org/archive/sipx-users/
>
_______________________________________________
sipx-users mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-users/

Reply via email to