|
We have 3com SuperStack III switches here that don't respect the ARP updates
that heartbeat/fake sends out. This was very frustrating when we were doing
testing of our heartbeat system. We still haven't taken the time to figure
out exactly what we need to change on them to make this work properly.
Right now if the primary doesn't completely fail(meaning it still has power
and the MAC address still is thought to be "alive" by the switch), the only
way to get the backup working properly is to turn off the primary. At that
point, the switch figures out that it has lost its link partner and
immediately(doesn't even take 90 seconds) starts using the backup. Very
strange indeed... has anyone else seen something like this? I imagine we
just need to change a setting on the switch.
-----Original Message-----
From: mack@xxxxxxxxxxxx [mailto:mack@xxxxxxxxxxxx]On Behalf Of Joseph
Mack
Sent: Wednesday, September 20, 2000 7:26 AM
To: ???
Cc: Joseph Mack; lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Subject: Re: sending arp ...
??? wrote:
>
> Thank you for your explaination.
> I have another question.
>
> > 1. do nothing: the arp table for ethernet expires in about 90secs and
> > the router will broadcast the request "who has VIP tell router" and
> > the new director will reply.
>
> Is the arp request broadcasted and reached every machine in the network ?
yes (I assume your netmask/broadcast addresses are set properly on all
machines)
> Isn't there any device, switching hub for example, that intercept that
> request ?
routers won't allow the request to be passed to another network, but this is
not
your problem. As far as I know a switch should pass arp packets.
> I'm testing HA in an environment where two directors are under a switching
> hub.
> The arp table of the router seems to be never updated after IP takeover
take
> place.
> (Here I didn't do send_arp)
> I couldn't connect to the VIP for more than an hour.
how does the entry for the first VIP get into router initially?
> Even though sending arp replies can be a solution, I wonder why the arp
> table
> didn't updated for so long a time.
> I suspected the switching hub to be the cause of this problem, but I have
no
> idea.
I have no idea either.
random suggestions:
1. Is the router set to NOT accept arp updates (for security reasons say)
2. Can you connect the new director to the router directly (ie no switch)
and then try to ping the VIP from the router and then look at the router's
arp table. The swap the VIP to the first machine and see if the router's
arp table gets updated. Do it again with the switch in line.
Joe
--
Joseph Mack PhD, Senior Systems Engineer, Lockheed Martin
contractor to the National Environmental Supercomputer Center,
mailto:mack.joseph@xxxxxxx ph# 919-541-0007, RTP, NC, USA
|
