|
Ok lets start again...
2 LVS servers both connected into a on ETH0 Cisco 2924xl which also has our
upstream connection plugged into it. ETH1 of the LVS's are on 192.168.0.x IP
addresses and are connected into a backend Cisco switch.
We have 2 web servers sitting on 192.168.0.x IP addresses also plugged into
the backend switch.
Everything works fine expect when I simulate the LVS service dying on the
primary LVS. Since I'm having to do all this remotely all I can do is drop
the LVS service which kills the heartbeat port. As such the backup LVS can't
connect to the primary LVS on the hearbeat port so assumes it's down. So it
brings up ETH0:1 (VIP) and ETH1:1 (Default GW for the 192.168.0.x subnet).
Locally within the subnet the servers are on it seems to have failed over
ok, web pages can be requested and are served by the backend web servers.
The frontend Cisco switch sees the VIP has moved ports and is now on a
different server.
However our upstream doesn't see the switch over it still thinks the primary
LVS is serving the VIP. I'm guessing this is because the MAC address the VIP
was on is still live (ETH0).
So I expect that the primary LVS has to actually die so that ETH0 is not
visible anymore. Since it wouldn't be visible the MAC address would
disappear from ARP and once the upstreams router sees this change then it
should all work ok. I'm gonna have to go to site to test this though.
My concern is if the LVS service does encounter a fault and dies on the
primary LVS then web requests will still fail since our upstreams router
will still see the MAC address of the primary LVS. And since the last it
knew the VIP was on that MAC address and that MAC address is still reachable
it isn't going to send any requests to the backup LVS.
Tony
-----Original Message-----
From: lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx
[mailto:lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx] On Behalf Of Joseph Mack
NA3T
Sent: 11 March 2006 18:06
To: LinuxVirtualServer.org users mailing list.
Subject: RE: IP Update - maybe an arp issue
On Sat, 11 Mar 2006, Tony Spencer wrote:
> I think my problem is the same as the first entry in 5.18.
>
> I think because the primary ETH0 is still up although I've failed over the
> LVS by stopping the service
which service?
> on the primary LVS, and since the heartbeat
> socket disappears the backup LVS brings up the VIP.
>
> The backup LVS logs show:
>
> Mar 10 14:24:17 lvs2 kernel: send_arp uses obsolete (PF_INET,SOCK_PACKET)
> Mar 10 14:24:22 lvs2 pulse[15503]: gratuitous lvs arps finished
> Mar 11 08:34:39 lvs2 pulse[18224]: gratuitous lvs arps finished
> Mar 11 08:36:40 lvs2 pulse[18349]: gratuitous lvs arps finished
>
>
> So I think it would be fine if the primary LVS actually died
what does died mean?
> and took it's ETH0 IP with it
the directors should have non-VIP IPs on the outside (and
inside)
> but in the event the LVS service
there is no LVS service. What do you mean here?
> just died leaving ETH0
> still up it won't failover correctly.
what's "it"?
I'm lost here. Can you start again?
Joe
--
Joseph Mack NA3T EME(B,D), FM05lw North Carolina
jmack (at) wm7d (dot) net - azimuthal equidistant map
generator at http://www.wm7d.net/azproj.shtml
Homepage http://www.austintek.com/ It's GNU/Linux!
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
|
