|
Ramprasad A Padmanabhan :
>
> I dont want to complicate my job , more than it is required. ( It is
> already in a mess ). I am kind of stuck implementing LVS.
It was the same for me at the beginning...
> Since yesterday I have been reading the docs LVS Howto, on setting up a
> LVS. I have decide to use Direct routing.
It is a good choice.
> Suppose I use Linux kernel 2.6.11.x ( SUSE ) as the LVS director and
> the realservers as Redhat machines , I want to know if
>
> 1) Should I disable arp on lo , only on the director machine.
You should disable ARP on the realservers NOT on the director. On the
realservers, just use :
net.ipv4.conf.lo.arp_ignore = 1
net.ipv4.conf.lo.arp_announce = 2
net.ipv4.conf.all.arp_ignore = 1
net.ipv4.conf.all.arp_announce = 2
> 2) If all my machines ( LVS director + realservers ) have valid
> internet IP's do I require to have a private network between my director
> and realservers.
It is a better solution but it works even with only one network. I use it
another way, I have DR setups that are only on my private network (for
internal application backends). So, if yuoou do not mess your ARP
configuration, it is possible to have all your LVS on your public network.
BUT, it is not wise to have administration flows on the same network as
internet flows, so, I would advise you to route your queries from director to
realservers via the private network.
Say :
Query -> eth0 on director (public (VIP) -> eth1 on director (private RIP) ->
eth1 on realserver (private) -> lo:x on realserver (VIP) -> APPLICATION ->
lo:x on realserver (outgoing paquet has VIP as the source address) -> eth0 on
realserver (public RIP) -> client.
> I know these questions are too basic , but for a newbie the docs are a
> lot of mumbo-jumbo. I will be reading them again , but I can sure use
> help.
Yes you can. It is not so easy to setup the first LVS. But when you have one
running, you need to re-read the docs again :).
François.
|
