On Thu, 6 Nov 2008, [EMAIL PROTECTED] wrote:

Thank you for the explanation you gave about anycast.
Heartbeat doesn't seem to implement this feature.
Do i have to conclude that it is simply impossible to build a cluster on two
different subnets with Heartbeat?

I mentioned anycast simply because it might just have been a possibility for handling a single IP address across multiple subets (or even nets) if you already had it in your network.

But if the word "anycast" is new to you, then you probably do not have it available within your network. That is a discussion (and probably a deep and detailed discussion) that you would need to have with your local network people.

So (unless your network people are very enthusiastic) then it is probably best to forget that I ever mentioned it.



To configure heartbeat to move an IP address between nodes, the network needs to provide the capability to support that IP address wherever it might be across the relevant subnets.


So this question really about network routing, not about heartbeat.

------------------
For a moment forget about heartbeat. Imagine you have two boxes on different subnets. And imagine that some remote client can ping the main IP address of those two boxes on their different subnets.

1. Imagine (manually) choosing and configuring a third (potentially floating) IP address on the first box.

2. Imagine pinging that IP address. Presumably it would work. But this is probably because you chose that IP address so it would work on the first box; that is, an IP address that belongs to the first subnet.

3. Imagine (manually) shutting down that IP address on the first box, and (manually) configuring it onto your second box.

4. (Here's the sticking point.) Imagine pinging that IP address. It probably won't work, because that IP address is not reachable (routable) from the main network. That is because the IP address was chosen to be on the first subnet, but is now housed on the second subnet; the network routing is pushing the ping the wrong way.

If that is the case (and it probably IS the case), then it is a feature of your network that won't let it reach that IP address, which logically belongs on one subnet, when it is physically located on the second subnet.
------------------


(Now, returning to heartbeat, which we had temporarily laid aside...)

In this regard, heartbeat is simply an application handling network packets, and it is subject to the routing behaviour offered (or not offered) by your network.

And heartbeat would not need to "implement anycast" (a concept which actually doesn't have any real meaning) because anycast is purely a network routing feature which, from heartbeat's persepctive (as an application sitting on the network) would be "unicast".

Hope that helps.



--

:  David Lee                                I.T. Service          :
:  Senior Systems Programmer                Computer Centre       :
:  UNIX Team Leader                         Durham University     :
:                                           South Road            :
:  http://www.dur.ac.uk/t.d.lee/            Durham DH1 3LE        :
:  Phone: +44 191 334 2752                  U.K.                  :
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to