|
On Thu, Dec 11, 2003 at 07:12:32AM -0500, kakani@xxxxxxxxxxx wrote:
> Hi All,
>
> I am new to this list. Did explored the archive a bit to find some
> information that I am looking for but could not find it. So writing
> this mail..
>
> I've been evaluating high-availability solutions for one of my
> company's product. This product (a Wireless TCP Gateway - WTCP), a
> optimization of TCP for wireless links. The product is still in
> development phase. As a proof of concept I have to build a complete
> gateway with hardware, high availability, Hot stand by (maintaining
> the connection status), and load balancing.
>
> I've been looking at HP's HA service software and its features. Also I
> looked at some of the Cisco content switches which provide such
> features. Let me come straight to the questiones,
>
> 1. How is the fail-over signaling mechanism for LVS or related
> technologies. For HP HA service, it provide a hardware, application or
> network failure signal on which appropriate action can be taken. So
> the failure could be the network adapter, it could be application
> running on that server, or it could be the failure of some link on the
> path of server which blocks traffic to and fro from the server. As far
> as I can understand, the LVS (and the ultramonkey.org) provides
> heartbeat mechanism for failover signaling. I would like to know if it
> supports the hardware/device/application/link failure signaling?
No, heartbeat does not support, though it will in the future
(hopefully). At this stage it can monitor the status of other nodes
by sending heartbeat packets. It can also monitor what are called
quourum devices by pinging them. It is also possible to write API
modules to tell heartbeat to fail over. I guess this could be used,
to some extent, to monitor interfaces, aplications, whatever.
I would suggest asking questions about heartbeat of this nature
on the linux-ha lists which can be found at
http://www.linux-ha.org/contact/
> 2. I read some of the papers on ultramonkey.org which kind of talks
> about providing active-active (hot standby) with state maintainance. I
> am not sure if that is applicable to LVS also? Is is really possible
> to provide hot standby using LVS? By hot standby I mean when the
> active gateway fails, it should transfer the control to the standby
> server, and the standby server should be able to handle all the
> maintain the current TCP connections (so the client does not actually
> re-initiate connection).
Yes.
--
Horms
|
