Failover scenario, using DNS with A records and "floating" IP address

B

Brian K

I'm using (or misusing?) DNS entries for the purposes of a simple
failover environment.

I'm using OpenView Storage Mirroring to replicate to
directories full of miscellaneous files (over two terabytes) between two
identically equipped Windows 2000 servers.

The live servers I'll call ServerA and ServerB. ServerA has been in place
for about four years now, and is on a domain recently migrated from NT4 to
Win2K AD, mixed-mode. The domain HAS a WINS server active on it, but the
two servers themselves don't have the WINS server listed in their network
properties. The machines have static IP addresses. At one point in their
existence, there were static WINS mappings entered on the WINS server for
both the machines, but those are long gone.

Now, to the successful test environment:

In my test environment, I have two servers, which I'll call TestA and TestB.
For the illustration, I'll have TestA use IP address 192.168.1.10, and TestB
192.168.1.11.

I then renamed TestA to TestC, and manually entered TestA into DNS as a
Host(A) record, pointing to IP address 192.168.1.12 (which was unassigned to
any machine.)

At this point, I can add 192.168.1.12 to any machine on our network as a
secondary IP address, and that machine will respond as if it's TestA: "net
view", drive mapping, SQL connectivity, anything you can think of, works
successfully. I can remove that IP address from TestC, move it to TestB,
and TestB immediately begins responding as if it were TestA. I can assign
that IP address to my XP desktop machine, or 2K lab workstation, etc. etc.
and it seems to function just perfectly.

Now, back to my unsuccessful production environment:

I rename ServerA to ServerC, and manually enter ServerA into DNS as a
Host(A) record pointing to a new unused IP address. Just like my test
scenario.

I assign the new unuses IP address to ServerC, and receive "Error 5: Access
Denied" when attempting to use "net view" or "net use." SQL connectivity
fails with "access is denied or server does not exist" for approximately 85%
of my desktop machines, but some oddly still function properly.

nbtstat -R, ipconfig /flushdns, ipconfig /registerdns, nbtstat -RR and a
reboot of the server and client PC's has no effect on the symptoms.

I CAN get the error message to go away and connect to ServerA if I map a
network drive and select the "Connect using a different username option" in
the "map network drive" dialog. I can then enter alternate credentials, and
as long as the network drive is mapped the problem is solved. The moment I
unmap that network drive, however, problem recurrs.

Also, I can even enter the already-logged-on user into the alternate
credentials dialog, and it will fix the problem as well, even though it's
the exact same username/password as is physically logged on to the machine.

However, to FURTHER compound the problem, I can take the IP address or
TestA, assign it as a secondary Ip address to ServerC, and everything works
great as far as TestA is concerned. The ServerA Ip address causes the exact
same symptoms no matter where it's assigned; any machine that has the
ServerA IP address responds with "access denied" to a "net view" command.

Is that confusing enough for everyone?

What I'm ultimately going to do is try with a cname record for ServerA to
point to ServerC, rather than trying the ip address/A record trickery, but
the manager wants a number of people to be able to fail the server over that
I'm quite frankly scared to let log on to the DNS server to make the change
to the cname; I'd rather just let them add an IP address to a server, rather
than go mucking about on our DNS server.

Brian K
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Top