|
Hi,
>
> Hello,
>
> On Thu, 18 May 2000, Madhav wrote:
>
> > Hi,
> >
> > > On Wed, May 17, 2000 at 03:40:17PM +0530, Madhav wrote:
> > > > Hi all,
> > > > When I am trying to connect to my LVS-NAT VIP 172.16.31.71 from
> > client
> > > > with ip 172.16.1.4( it was in a diff. floor and so it has to go
through
> > main
> > > > switch), it was taking a long time when compared with the time it
takes
> > to
> > > > connect to the VIP from a client in the same sub net as LVS-NAT
> > cluster(e.g.
> > > > an address like172.16.31.74).
> > > > Here is my setup.I have used two directors for my
LVS-heartbeat-mon
> > > > setup: master-172.16.31.72, slave-172.16.31.73 and two real servers
with
> > > > 10.1.10.25 and 10.1.10.26.
> > > > I am guessing that RARP resolution is causing the delay, because
> > both
> > > > IPs 172.16.31.71(VIP for LVS) and 172.16.31.72(RIP for master
director)
> > are
> > > > claiming the same ethernet address( when the VIP is attached to the
> > master
> > > > director) . If this is the reason how can I reduce this? and if not
> > please
> > > > tell me what is causing this delay.
> > >
> > > You have two machines on the network with the same IP address?
> >
> > NO NO.. A machine has two IP addresses.
> > I should have said ARP resolution I suppose. I am not very sure. Anyway
here
> > is the senario. The the address 172.16.31.72 is the eth0 address of my
> > director1 or master and 172.16.31.71 is the eth0 alias of the same
machine.
> > So the director machine has two IPs. That is what it should be in
LVS.right?
> > I have installed a qmail server in both the real servers over NFS. LVS
is
> > providing load balance for mail server.
> > When I connect to the LVS VIP through netscape from a client, which
is
> > in the same subnet as that of LVS, it is connecting immediately. But
when I
> > am trying to connect from a client in another subnet, whose IP is
> > 172.16.1.4, it is taking a very long time( nearly 30 seconds). How can I
> > reduce this delay when the client is in a different subnet. Is there any
way
> > we can hide the eth0 address. Thst solves the problem I guess.
>
> If this is only a delay what about DNS or IDENT problems?
> Could be that the IDENT protocol used from the mailers causing
> the problems? If Yes, You have to stop using IDENT. The LVS >HOWTO
> includes information about the IDENT problems. If this is not
I just checked the LVS-HOWTO it is a very big doc. I am going to look into
it. Thanks.
> the problem you have to show us a picture of your devices, IPs,
> routes, etc.
Here is the picture of my LVS-NAT
-----------
| |
| client | if client IP is 172.16.1.4 then it
is very slow
|________| if it is 172.16.31.74 then it is very
fast.
|
|
Virtual IP(172.16.31.71) (eth0 alias)
___________________________________________________________________ LAN
____|_____
_____|_____
| | (director has 2 NICs)
| | (director has 2 NICs)
| director1 | eth0:172.16.31.72 |
director2 | eth0:172.16.31.73
|__________| eth1:10.1.10.100
|__________| eth1:10.1.10.50
|
|
|
|
|
|
HUB ----------------------------------------------------------------------
| |
| |
| |
____________ ____________
| | | |
| realserver1 | | realserver2 |
|____________| |____________|
10.1.10.25 10.1.10.26
Hope this helps to help me.
Thanks,
Madhav
|
