|
Hi, ratz
On Tue, Nov 14, 2000 at 02:26:02PM +0100, ratz wrote:
> Hi Wenzhuo,
>
> > I suppose it should be:
> >
> > route delete <VIP's network> mask <VIP's netmask>
>
> Makes definitely more sense ;) Joe, could you change this
> please? Anyway it is not needed since it works without
> deleting this route. The few NT systems I had to setup, I
> never deleted this route.
>
> > > This will make the packets destined for this network will
> > > go through the other network interface, not this MS Loopback
> > > interface.
> > >
> > > As I remember, setting its netmask to 255.0.0.0 also works.
> > >
> > It works, but hosts inside the VIP's subnet will not be able to
> > connect to the real server. If you've got any host inside the vip's
> > subnet (not matter how small it is), you'll have to delete its subnet
> > route.
>
I reached this conclusion when I was setting up a 2-node NT (SP5 and
SP6, Chinese version) cluster a week ago. After a couple more of
tests, I have to say this is only partially correct (i.e. false ;).
Now I'd like to modify my previous statement a bit. :)
Let's denote RIP's subnet as A, and VIP's subnet as B.
If A > B, hosts in B will not be able to connect to the real servers
(RIPs of course).
If A = B, subnet route A takes precedence, so it works fine.
If A < B, hosts in B but not in A will not be able to connect to the
realservers.
Setting VIP's subnet to the same as that of the RIPs seems to be the
ideal choice, in which case, RIP's subnet route takes precedence.
Anyway, MS NT is confusing.
Wenzhuo
> Are you sure? Maybe I missunderstood you statement or NT's routing
> table parsing or I simply cannot follow you. As far as I tested it
> (although mostly with different Unices) I had no problem connecting
> to the realserver within the netmask of the VIP. However you cannot
> connect to the VIP within the cluster but this is a different cup
> of beer.
>
> Best regards,
> Roberto Nibali, ratz
>
> --
> mailto: `echo NrOatSz@xxxxxxxxx | sed 's/[NOSPAM]//g'`
>
|
