2 NICs different IPs in same network?

[RQ]

Hi,
Can 2 NICs, different IPs work in same network? Could be error from machine
or from the network (ethernet)?
If not, why?
Thanks,
Roger

 

[Phillip Windell]

Put the IP#s on the same NIC. Use one Nic. Why do you think your need two
nics with two IP# from the same subnet?

 

[RQ]

Thanks Phillip,
I know this is not a good idea, but our database guy wants to set second NIC
up on secondary database server using primary server's IP when the primary
goes down and still keep secondary server's first NIC IP active.
I don't know how to prove this is a bad idea. Please advice.
Roger

Put the IP#s on the same NIC. Use one Nic. Why do you think your need two
nics with two IP# from the same subnet?

--

Phillip Windell [MCP, MVP, CCNA]
www.wandtv.com

Hi,
Can 2 NICs, different IPs work in same network? Could be error from machine
or from the network (ethernet)?
If not, why?
Thanks,
Roger

 

[Ace Fekay [MVP]]

In

Thanks Phillip,
I know this is not a good idea, but our database guy wants to set
second NIC up on secondary database server using primary server's IP
when the primary goes down and still keep secondary server's first
NIC IP active.
I don't know how to prove this is a bad idea. Please advice.
Roger

Phillip,

This was answered in the DNS newsgroups with a great response as to why NOT
to do it by Kevin, unless the NICs are teamed.

--
Regards,
Ace

Please direct all replies ONLY to the Microsoft public newsgroups
so all can benefit.

This posting is provided "AS-IS" with no warranties or guarantees
and confers no rights.

Ace Fekay, MCSE 2000, MCSE+I, MCSA, MCT, MVP
Microsoft Windows MVP - Active Directory

HAM AND EGGS: A day's work for a chicken;
A lifetime commitment for a pig.

 

[RQ]

Phillip,
I'd like to have your answer about this issue.
Thanks,
Roger
"Ace Fekay [MVP]"

 

[Phillip Windell]

Thanks Phillip,
I know this is not a good idea, but our database guy wants to set second NIC
up on secondary database server using primary server's IP when the primary
goes down and still keep secondary server's first NIC IP active.
I don't know how to prove this is a bad idea. Please advice.
Roger

Good reason why database guys should be allowed to do networking.

1. You can't have two machines on the network with the same IP#.

2. The solution is extremely easy,...if the first server goes down for an
extended period, then add its IP# to the *one* nic on the second server as a
secondary address. Remove it before the first server is brought back up
again.

3. Another Solution could be to do Server Clustering with two identical DB
Servers.

 

[Phillip Windell]

Ok.

I don't watch that group and it must not have been crossposted so it didn't
show up here. Dunno if my reply was anything like theirs.

--

Phillip Windell [MCP, MVP, CCNA]
www.wandtv.com


"Ace Fekay [MVP]"

 

[RQ]

Thanks Pillips,
I suggested to do that way, but I'd like to know why 2-nic 2-IP same LAN is
not allowed and what could be happen in the network if apply this
configuration?
Roger

 

[Ace Fekay [MVP]]

In

Thanks Pillips,
I suggested to do that way, but I'd like to know why 2-nic 2-IP same
LAN is not allowed and what could be happen in the network if apply
this configuration?

My take on this is if you have two IPs with the same name, then there will
be a conflict with NetBIOS names (duplicate names, same IP). This is the
basis of MS networking that's difficult to circumvent. Many applications
rely on NetBIOS support. Assuming you will be wanting this for 'failover', I
can understand but to make two IPs work on the same network (same segment or
enterprise) you need to disable NetBIOS, MS CLient Service and the F&P
service on one of them. But considering its for a database, and IIRC, you
are using SQL, SQL requires NetBIOS name resolution, so this automatic
failover solution will fail. Exchange behaves the same way, since MAPI
clients REQUIRE NetBIOS resolution.

You can fudge this by trying two DNS A entries with the same name and give
it the two different IPs, but the issue comes down to, since Round Robin is
in effect, that the server will offer one IP to one client, and then the
other to the next query. If the one NIC is down, it will not stop DNS from
giving that IP to a querying client and hence the client cannot connect
since the NIC is down. This still does NOT address the duplicate name issue
with the two IPs for the same name. This also does not work with Subnet
Prioritization, since both IPs are on the same subnet.

As previously posted, its highly suggested to use Clustering to achieve this
effect. Or use NIC Teaming (based on the NIC manufacturer's software) that
gives both NICs the same IP and they work in conjunction. meant for load
balancing, but IIRC with teaming, that should one fail, the other continues
servicing, but warnings are giving. You would need to read up on the
vendor's product sheet for more info on this.

Now if I missed something else on this, Phillip, or anyone else that reads
this, please do add ro comment on this. I just want the poster to understand
the mechanisms behind this in order to realize what is occuring and why.

--
Regards,
Ace

Please direct all replies ONLY to the Microsoft public newsgroups
so all can benefit.

This posting is provided "AS-IS" with no warranties or guarantees
and confers no rights.

Ace Fekay, MCSE 2000, MCSE+I, MCSA, MCT, MVP
Microsoft Windows MVP - Active Directory

HAM AND EGGS: A day's work for a chicken;
A lifetime commitment for a pig.

 

[Phillip Windell]

Thanks Pillips,
I suggested to do that way, but I'd like to know why 2-nic 2-IP same LAN is
not allowed and what could be happen in the network if apply this
configuration?

It can't know which physical interface to use for traffic. The Routing Table
is used to make those decisions and it has no "criteria" to know which
interface is the right one to use. Look at this example, if you have a
single Int with a single IP of *.70 and a LAN Router at *.1 for a DFG you
have an entry like this:

Network Dest Netmask Gateway Interface
Metric
0.0.0.0 0.0.0.0 192.168.1.1 192.168.13.70
1
192.168.1.0 255.255.255.0 192.168.1.70 192.168.1.70 1
192.168.1.70 255.255.255.255 127.0.0.1 127.0.0.1 1

The First line is the Default Gateway pointing at 192.168.1.1,
The Second line is the Default Route for the local subnet,
The Third line is the Loopback route. If you add a second IP# like *.75 to
the *same* NIC you will get this:

Network Dest Netmask Gateway Interface
Metric
0.0.0.0 0.0.0.0 192.168.1.1 192.168.13.70
1
192.168.1.0 255.255.255.0 192.168.1.70 192.168.1.70 1
192.168.1.70 255.255.255.255 127.0.0.1 127.0.0.1 1
192.168.1.75 255.255.255.255 127.0.0.1 127.0.0.1 1

Notice now that there is a second Loopback Route for the .75 address but the
Default Route for the local subnet has not changed and the Default Gateway
has not changed. So it still knows how to send traffic out that Interface
without "blowing its brains out".

But now let's screw it up. With a second NIC in the machine using the *.75
address instead of putting in on the same NIC as before you get this:

Network Dest Netmask Gateway Interface
Metric
0.0.0.0 0.0.0.0 192.168.1.1 192.168.13.70
1
0.0.0.0 0.0.0.0 192.168.1.1 192.168.13.75
1
192.168.1.0 255.255.255.0 192.168.1.70 192.168.1.70 1
192.168.1.0 255.255.255.0 192.168.1.75 192.168.1.75 1
192.168.1.70 255.255.255.255 127.0.0.1 127.0.0.1 1
192.168.1.75 255.255.255.255 127.0.0.1 127.0.0.1 1

Whola! Now you have two Default Gatways with the same metric and two
Default Routes for the same local subnet with the same metric. There is
nothing to distiguishing them apart to determine which one to use for
traffic on it's own subnet,... or for "unspecified" routes which is the
Default Gateway's job.

Now it can go by the NIC with the highest prioity, or it can "flip a coin",
or it can "blow its brains out". Logically it seems it would go by the
"priority" of the Nic, but then sometimes it may "flip a coin", and other
times it just "blows its brains out".

The following article doesn't describe exactly this situation but it does
reveal the machines behavior patterns in these types of situations.

Microsoft Windows XP - Multihoming Considerations
http://www.microsoft.com/resources/documentation/windows/xp/all/reskit/en-us/prcc_tcp_qpzj.asp?

 

[Phillip Windell]

"Ace Fekay [MVP]"

Now if I missed something else on this, Phillip, or anyone else that reads
this, please do add ro comment on this. I just want the poster to understand
the mechanisms behind this in order to realize what is occuring and why.

I think you are right "on the money". I would add that the teaming wouldn't
work in his case because he is ultimately dealing with two machines with the
same IP#, on top of the fact the one machine has two IP#s with two NICs in
the same subnet where the second NIC has the same IP# that the other
physical machine is already using.

I think the idea is that the second NIC that duplicates the other machine's
IP# will somehow be "dormant" or deactivated unless the other machine fails.
If the other machine fails, then the "dormant" NIC on the first machine is
activated to allow the machine to act as a "failover machine" when the
original one dies. Obviously I don't that such a thing could ever be made
to work short of "server clustering".

If they are not clustered then the machines would have to be setup the same
and keep the data in the DB in "sync" between the two machines. Then if the
main one dies, it has to be unpluged and the IP# it was using would be added
to the "one and only" NIC on the "backup" machine so it could fill-in for
the main while it is being repaired. But by the time this is all setup and
the DBs are somehow rigged to stay in "sync" 24/7, about 80% of the
clustering requirements would already have been done, so they might as well
be clustered and forget it.

Things just work so much better when people perform tasks according to
industry standards & practices, instead of trying to "outsmart the system".
No offense to RQ since I understand that this is being pushed on you.

 

[RQ]

Pillips and Ace,
Thank you very much for spending time to asnwer my question.
I'll talk to my co-worker to find out a good solution for this issue.
Thanks again!
Roger

 

[Ace Fekay [MVP]]

In

Microsoft Windows XP - Multihoming Considerations

http://www.microsoft.com/resources/documentation/windows/xp/all/reskit/en-us/prcc_tcp_qpzj.asp?

Phillip, nice of you to go thru the trouble of discovering what happens in
both scenarios!!


Interesting article about XP. The first thing I see is that NetBIOS binds to
the first IP on EACH NIC. So this still holds true to the dupe name error.
Also, just to point out about the routing table, unless metrics are
specified, what I surmise is that with mutliple NICs, regardless of the
routing table, it will use the first in the binding order.

My feeling is that just to install the extra NIC and disable it. Manually
enable it when the other fails. The SEC folks do that (IIRC). Those guys
have 3 NICs in their machines. Triple redundancy. With transactions at 10's
of thousands/sec, you need to insure quick failover. Just disable the bad
NIC, change the wire and enable the next NIC in line.

--
Regards,
Ace

Please direct all replies ONLY to the Microsoft public newsgroups
so all can benefit.

This posting is provided "AS-IS" with no warranties or guarantees
and confers no rights.

Ace Fekay, MCSE 2000, MCSE+I, MCSA, MCT, MVP
Microsoft Windows MVP - Active Directory

HAM AND EGGS: A day's work for a chicken;
A lifetime commitment for a pig.

 

[Ace Fekay [MVP]]

In

Pillips and Ace,
Thank you very much for spending time to asnwer my question.
I'll talk to my co-worker to find out a good solution for this issue.
Thanks again!
Roger

Check out my latest response about what the SEC does.


--
Regards,
Ace

Please direct all replies ONLY to the Microsoft public newsgroups
so all can benefit.

This posting is provided "AS-IS" with no warranties or guarantees
and confers no rights.

Ace Fekay, MCSE 2000, MCSE+I, MCSA, MCT, MVP
Microsoft Windows MVP - Active Directory

HAM AND EGGS: A day's work for a chicken;
A lifetime commitment for a pig.

 

[Phillip Windell]

Phillip, nice of you to go thru the trouble of discovering what happens in
both scenarios!!

Sometimes I think I just have too much time on my hands...

of thousands/sec, you need to insure quick failover. Just disable the bad
NIC, change the wire and enable the next NIC in line.

I wondered if that would work ok. I wasn't sure if the Nic being there
effected the Routing Table (or not) even if it was disabled.

 

[Ace Fekay [MVP]]

In

I wondered if that would work ok. I wasn't sure if the Nic being
there effected the Routing Table (or not) even if it was disabled.

I believe if its disabled, the system calculates the routing table without
it, since its not an active resource.


--
Regards,
Ace

Please direct all replies ONLY to the Microsoft public newsgroups
so all can benefit.

This posting is provided "AS-IS" with no warranties or guarantees
and confers no rights.

Ace Fekay, MCSE 2000, MCSE+I, MCSA, MCT, MVP
Microsoft Windows MVP - Active Directory

HAM AND EGGS: A day's work for a chicken;
A lifetime commitment for a pig.