|
On Tue, May 09, 2000 at 09:57:28AM -0400, Paul Lussier wrote:
>
> In a message dated: Mon, 08 May 2000 21:58:47 PDT
> Horms said:
>
> >Actually, to do this with Ultra Monkey you would probably use heartbeat.
> >Ldirectord is used to monitor the servers that are available when load
> >balancing is done and will include all available nodes in the pool of
> >servers to which connections can be laod balanced. Heartbeat on the other
> >hand can be used to manage 2 or more servers that own particular resources
> >that are available on exactly one node at any given time.
>
> Okay, let me see if I have this completely understood :)
>
> - LVS uses Ldirectord for load balancing among the "real servers"
LVS does the load balancing. However LVS has no conccept of failures,
much in the same way that a static routing table constucted using
/bin/route doesn't adapt to network failures. Ldirectord may be
used to monitor the servers that the LVS host load balances to
and take them in and out of the LVS table (using ipvsadm) as
required.
> - heartbeat is used by systems configured in a redundant/failover
> configuration to monitor each other
>
>
> So, for the failover side of LVS, I need
>
> > - ipvs kernel patch
> > - mon/fping
> > - Period.pm
>
> - heartbeat
>
> Do I need anything else, like Fake, or Failoverd? Or are these
> funtionalities accounted for in heartbeat?
Heartbeat should cover the functionality of these.
--
Horms
|
