Re: haproxy listening to all interface is safe for a heartbeat cluster? (was: High Availability question)

Hi Dirk,

We use heartbeat with haproxy and don't have any issues. Our setup is pretty simple, we have a VIP that floats between the two servers by heartbeat and when I move it over via hb_takeover or hb_standby (we haven't had a failover happen besides my manually moving it or rebooting the primary to force a fail over) it takes maybe a second at the most for the traffic to start going through that server.

Since the apache servers are responding to the internal IP of the haproxy server when the secondary takes over and starts sending the requests the apache servers just start sending to that servers IP address.

I haven't seen any issues with this setup and have had great results so far. I don't see why your would force a start on apache unless you have them residing on the same server as haproxy.

Anyway wanted to respond and let you know that we have had no issues with heartbeat failover between haproxy servers.

Joe
**

Dirk Taggesell wrote:
> Hi all,
>
> having read the previous tread, what I understood was that for a
> heartbeat cluster with haproxy it is a safe configuration to have
> haproxy just listen to all interfaces and it will automatically pick up
> the newly assigned address once a fail-over happens?
>
> Can I be sure haproxy recognises the new IP address after a fail-over
> without it being restarted (or may be this is nothing haproxy has to
> worry about, because it's a kernel thing)?
>
> How about may be Apache?
>
> Until now I always created orders and colocations to start and stop
> haproxy/apache when a fail-over takes place - just to be correct.
>
> --
> kind regards
> Dirk Taggesell
>
> --
> This message has been scanned for viruses and
> dangerous content by *MailScanner* <http://www.mailscanner.info/>, and is
> believed to be clean.

-- 
This message has been scanned for viruses by Colocube's AV Scanner