OK, You guys are pretty smart. This is not LINUX, but Networ

Main Page | Report this Page

Linux Forum Index » Linux - Suse Forum » OK, You guys are pretty smart. This is not LINUX, but Networ

Page 1 of 1

OK, You guys are pretty smart. This is not LINUX, but Networ

Author

Message

Sitting Duck

Posted: Thu Nov 09, 2006 1:49 pm

Guest

We have a very odd problem here. After our mail server went down, we
offloaded our mail to off-site, but since we have been having trouble with
connections.

We are running a 100-130 node network with a PDC and a BDC both running
Windows 2003 Server. Our mail system was a SCO Unix Open Server 5.0.4
server that was NOT doing any mule work for the network/internet.

Our workstations are a mish-mash of M$ OS's, from NT to 98 to W2000 to XP.
This problem affects all the workstations like a rolling blackout.

The issue is, some machines are unable to connect to the internet/e-mail.
The perplexing part: all machines can see the network, can see the
servers, and can see shares on other machines. All machines can see the
Intranet page we have set up inside the company on the PDC. All machines
can Ping, all machines have DNS resolution working, and all machines can
tracert to external servers like Google. The problem is, some machines
cannot access anything outside the domain. They can ping google, they can
tracert google, but if you try to browse google with a browser, the DNS
resolves and nothing happens. The same for e-mail. It can connect to the
e-mail server, it sees the e-mail server, but nothing gets sent or
received. We have tried different switches and routers, etc, and the
machines that have this problem are hosed.

But, a machine that is working perfectly fine will all of a sudden go
down, and then come back up anywhere from 15 mins to an hour or more.
We have doen virus scans, etc. This is a real odd problem.

Any ideas?

Chris Cox

Posted: Thu Nov 09, 2006 2:19 pm

Guest

Sitting Duck wrote:

Quote:

Way too many variables. Could be almost anything. I think you need
to plan some down time (maybe over a weekend) where you can take
apart the infrastructure and re-add elements piece by piece.

I know I've seen problems similar to this in a situation where a
company had some bad cabling.... and of course, we just "thought"
we were seeing what we "thought" we were seeing... very confusing...

If you have some network monitoring devices that can be put into
place... now it a good time to look at that as well.

BearItAll

Posted: Thu Nov 09, 2006 3:03 pm

Guest

Sitting Duck wrote:

Quote:

Are you sure that your email server was only serving emails? Could it have
had all or part of the dhcp and name resolution job also?

Because dns is so important to mail servers some IT hand all that side of
the job over to them, maybe letting it share with a proxy (i,e, squid) when
you are not using transparent proxies.

Also, if another device/server is serving dhcp, should the information that
comes over with the IP need changing? For example the gateway or proxy.
DHCP can pass a great deal of information to clients, could your
predecessor have made use of that?

Are the machines all in the same IP class, again pointing at the dhcp. As in
mine, machines that are not in the network class can not get out onto the
internet.

The machines going up and down could be requesting an IP from a dhcp server,
not finding it so they self allocate. I am not sure what the policy is for
MS self allocation, but presumably they all allocate in one of the two main
classes for that, which would allow them to see each other, while still
being out of class as far as internet access is concerned.

Basically, I guess I'm saying *dhcp*, but if you are having name problems
then a knockon is that the name resolving is shot, probably for the same
reason, i.e. machines are not pointing to or have no access to the right
place for name resolution.

Sitting Duck

Posted: Thu Nov 09, 2006 3:09 pm

Guest

On Thu, 09 Nov 2006 10:19:49 -0600, Chris Cox wrote:

Quote:

Sitting Duck wrote:
We have a very odd problem here. After our mail server went down, we
offloaded our mail to off-site, but since we have been having trouble with
connections.

We are running a 100-130 node network with a PDC and a BDC both running
Windows 2003 Server. Our mail system was a SCO Unix Open Server 5.0.4
server that was NOT doing any mule work for the network/internet.

Our workstations are a mish-mash of M$ OS's, from NT to 98 to W2000 to XP.
This problem affects all the workstations like a rolling blackout.

The issue is, some machines are unable to connect to the internet/e-mail.
The perplexing part: all machines can see the network, can see the
servers, and can see shares on other machines. All machines can see the
Intranet page we have set up inside the company on the PDC. All machines
can Ping, all machines have DNS resolution working, and all machines can
tracert to external servers like Google. The problem is, some machines
cannot access anything outside the domain. They can ping google, they can
tracert google, but if you try to browse google with a browser, the DNS
resolves and nothing happens. The same for e-mail. It can connect to the
e-mail server, it sees the e-mail server, but nothing gets sent or
received. We have tried different switches and routers, etc, and the
machines that have this problem are hosed.

But, a machine that is working perfectly fine will all of a sudden go
down, and then come back up anywhere from 15 mins to an hour or more.
We have doen virus scans, etc. This is a real odd problem.

Any ideas?

Way too many variables. Could be almost anything. I think you need
to plan some down time (maybe over a weekend) where you can take
apart the infrastructure and re-add elements piece by piece.

I know I've seen problems similar to this in a situation where a
company had some bad cabling.... and of course, we just "thought"
we were seeing what we "thought" we were seeing... very confusing...

If you have some network monitoring devices that can be put into
place... now it a good time to look at that as well.

Thanks for the input. But, sometimes the things you overlook are the
simplest ones of all!

We rebooted the router/firewall.
We didn THINK that was the problem because 80% of the people had
connectivity! Oh, well, live and learn.

And as far as monitoring equipment: I can't even get a cable tester! :(

One of the machines in question did have a bad cable (intermittant)

Thanks for the reply!

Bob Bob

Posted: Thu Nov 16, 2006 9:00 pm

Guest

Okay so lets get rid of a few possibilities.

- Not a routing issue as ping works
- Not a proxy cache/enough CAL's issue as email is also broken.
- Not an M$ update issue as you have various clients.

And

- What about packet length? MTU problems maybe as it crosses from your
private address space to Internet (DNS and ping are short packets) Do
longer ping packets work okay (ie 1500 bytes)
- What else did the SCO box do that may have helped the network load?
Was it a proxy cache or secondary DNS or DHCPD?
- Max numbers of TCP connections on the Internet side. (ping is ICMP and
DNS normally UDP) How about masquerade port capacity. They have a
timeout don't they? If you stopped using a proxy cache there would be a
lot of individual masquerading going on. Are clients seeing the proxy
config in W2003's DHCPD etc?

Have you swapped out the router/masquerade box at the firewall?

Flaky ideas but I had to say them!

Sorry for being no help!

Cheers Bob

Sitting Duck wrote:

Quote:

We have a very odd problem here. After our mail server went down, we
offloaded our mail to off-site, but since we have been having trouble with
connections.

Page 1 of 1

All times are GMT - 5 Hours The time now is Mon Nov 23, 2009 8:36 am Linux Forum Index