Re: failover slave doesn't send gratuitous arp

Janet Houser wrote:
> Hi folks,
>
> I was hoping you could assist me in solving a problem.  I'm configuring
> two failover servers following the guidelines for a High Availability: Single
> Virtual Service (see http://www.ultramonkey.org/3/topologies/ha-eg.html
> ).
>
> My system is running CentOS release 4.3 (Final) ( 2.6.9-34.ELsmp #1 SMP Wed Mar
> 8 00:27:03 CST 2006 i686 i686 i386 GNU/Linux)
>
> I have connected a crossover cable between the servers on eth1.  
>
> My ha.cf file on both servers looks like:
>
> debugfile /var/log/ha-debug
> logfile /var/log/ha-log
> logfacility     local0
> keepalive 2
> deadtime 30
> warntime 10
> udpport 694
> initdead 120
> udp     eth1            # Linux
>   
The row above seams wrong.
You want to say if you would like to use broadcast/multicast/unicast etc 
and which interface
to use for heartbeat messages. Have a look at 
http://www.ultramonkey.org/3/topologies/config/ha/single/real-server/ha.cf 
again.
(It is _highly_ recommended to use two interfaces for the heartbeat 
messaging)
> auto_failback off
> node    master.mydomain.com
> node    slave.mydomain.com
> debug 1
>
> My haresources on both nodes looks like:
>
> master.mydomain.com  \
>         IPaddr2::xxx.xxx.xx.xxx/21/eth0/xxx.xxx.xx.xx vsftpd
>
> I then start hearbeat (service heartbeat start) on the slave
> and then the master.  I notice that a gratuitous arp goes out
> from each of the servers as the service starts.  Since I start
> my master last, the arp tables contain the MAC address for this
> node and ftp works.
>   
Yes, in this case true since your heartbeat messaging does not work and 
your ha-servers dont know about each other.
If the would the prefered master (according to haresources) would start 
the service, the slave should standby.
> However, when I test failover by stopping (service heartbeat
> stop or /etc/init.d/heartbeat stop) on the master node, the slave
> does not send out a gratuitous arp.
>   
The slave should not have tried to bring up the service ip-address yet 
(if your hearbeat config is righet), but should do so now when it 
detects the master have failed/stopped.
The slave should send out a gratuitous arp just when it brings up the 
service ip-address.

> Since these machine sit on the DMZ of a FW, none of my machines
> can reach it until the FW arp cache clears.  Which, right now,
> is set for 4 hours.  I did set the FW cache timeout to 5 minutes
> which worked but I rather not change my FW.
>
> Likewise, I have a few machines sitting on the same LAN as the
> failover pair and I have to wait a couple minutes before the 
> workstation arp cache clears before I can connect to the slave
> server.
>
> If I send out (by hand) a gratuitous arp via:
>
> /usr/lib/heartbeat/send_arp -i 200 -r 5 -p
> /var/lib/heartbeat/rsctmp/send_arp/send_arp-xxx.xxx.xx.xxx eth0 xxx.xxx.xx.xxx
> auto xxx.xxx.xx.xxx fffffff
> fffff
>
> on the slave machine.  the LAN workstations and the FW update and all
> works great!  However, I can't seem to figure out how to force this 
> when the slave senses the heartbeat failure of the master.
>
> NOTE:  the X'x in the above IP addresses are of course the ip of the 
>        virtual interface that gets created on eth0:0 which is separate
>        from the ip's on eth0 and eth1.
>
> Any suggestions/advice would be grately appreciated!
>
>   
Hope this helps you...

/Mattias
> Cheers,
>
> Janet
>
>
> P.S.  This is a second post of this.  Don't think the first one made it into
>       the forum.  
>
>
>   


-- 
Ultra Monkey - http://www.ultramonkey.org/
To UNSUBSCRIBE, email to [email protected], with a body:
unsubscribe ultramonkey-users [email protected]
where "[email protected]" is YOUR email address.