Re: nfs and virtual interfaces
From: Darren Dunham (ddunham@redwood.taos.com)
Date: 04/19/03
- Next message: Urs Krolzig: "Re: IP accounting on per user base"
- Previous message: Jonas Bygdén: "Re: problem with setting up a jumpstart server."
- Maybe in reply to: Dan Stromberg: "nfs and virtual interfaces"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
From: Darren Dunham <ddunham@redwood.taos.com> Date: Sat, 19 Apr 2003 17:40:53 GMT
Dan Stromberg <strombrg@dcs.nac.uci.edu> wrote:
>> Dan Stromberg <strombrg@tesuji.nac.uci.edu> wrote:
>>> If I have 1 NFS server, with 1 normal interface and one virtual
>>> interface, can I have two different NFS clients NFS mounting from the
>>> two different interfaces?
>>
>> Sure.
> Forgive me for looking a gift horse in the mouth, but may I ask: Do you
> have experience with this, or are you guessing?
I can't recall doing this particular thing exactly, but I'm quite
certain of what the default NFS server and clients will do here.
Call it a suggestion that you should test for yourself... :-)
> If I have two names for one interface, would that not mean two A's and one
> PTR? ISTR that Sun NFS requires the A and PTR to match up (kind of like
> tcpd), making the 2nd A less useful for this purpose. I also STR having
> trouble with NFS mounting a CNAME.
The only side doing authentication is the server, not the client. The
client won't care a bit about the missing PTR.
>>> I'm worried that the machine is going to try to send all its return
>>> packets on one of the interfaces, irrespective of which interface a
>>> packet was received on.
>>
>> Wait, why do you care? It will send all the return packes on one of the
>> interfaces, but why does that affect what you're attempting to do?
> ISTR that Sun NFS, for security reasons, is particular about getting
> packets from an address it expected - as a limited for of authentication.
It does not.
Just to test, I took a NFS server, and ifconfiged a new virtual
interface.
server# ifconfig hme0:1 plumb 192.168.1.205 netmask + broadcast + up
server#
Then I mounted that virtual interface on a client in the subnet.
client# mount 192.168.1.205:/export2 /mnt
client# ls /mnt
local.cshrc local.profile san test
local.login plato tan
If you are doing default solaris mounts, you will probably do TCP
mounts. Since they are connection oriented, the specific IP address of
the named server will be used for incoming and outgoing connections
(although the interface used will be determined by the routing table).
I've tested the above mount. When I do a default mount, snoop shows the
NFS traffic from the server has a source IP address of 192.168.1.205.
When I force a UDP mount, the NFS traffic from the server has a source
IP address that matches the IP on hme0. Both mounts work.
There should be no issues here.
-- Darren Dunham ddunham@taos.com Unix System Administrator Taos - The SysAdmin Company Got some Dr Pepper? San Francisco, CA bay area < This line left intentionally blank to confuse you. >
- Next message: Urs Krolzig: "Re: IP accounting on per user base"
- Previous message: Jonas Bygdén: "Re: problem with setting up a jumpstart server."
- Maybe in reply to: Dan Stromberg: "nfs and virtual interfaces"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|
|
