On the work network I administer I had a fiber optic
module failure on one of my switches. I swapped it out
and restored connectivity over the network, but all except
about 10 users (out of 100) were able to connect to my
linux box. I was only able to get those 10 to connect by
assigning a static IP instead of using the assigned DHCP
address (the static one I assigned was in the same subnet,
utilizing the same dns and gateway as the ones assigned
by the DHCP server). I am at a total loss as to why this is.
The linux box is plugged into the same switch as a few
other servers (running w2k) and the users were all able to
connect to the windows machines. The only out of the
ordinary thing I can think of is that the internal network is
using a 192.168.x.x range with a subnet of 255.255.0.0
(was set up like this before I got there but has been
working fine with everything, including the linux box, for
2 years now). The only thing I changed was the fiber optic
module in the switch. Would greatly appreciate any
suggestions people might have. The linux box is running
a red hat but I'm not sure which version.

I was only able to get those 10 to connect by assigning a static IP instead of using the assigned DHCP address (the static one I assigned was in the same subnet,

Static IP assigned to what, the linux box or the 10 clients?

The linux box is plugged into the same switch as a few other servers (running w2k) and the users were all able to connect to the windows machines

Is the linux box a server or a client? If a server, and you're allowing it to get it's IP info from DHCP the question remains....why? Server's should always have static IP's so there's no chance of it ever changing.

The only thing I changed was the fiber optic module in the switch.

That shouldn't change anything unless you changed the switch configuration after swapping out the bad GBIC. Did you change the config on the switch? Is the configuration the same? Did you check?

The linux box is running
a red hat but I'm not sure which version.

Shouldn't matter what version but it's easy enough to find out. Open a terminal window and type uname -a

the linux box is a server and, of course, it has a static address assigned (192.168.0.x). The dhcp range given to the clients is 192.168.2.x, mask 255.255.0.0, gw 192.168.0.1. For these 10 users if they use their dhcp address they cannot even ping the address of the linx box. If I give a static addres (192.168.3.x - for example - same mask and gw) they can ping and connect to the services running on that box no problem.

I did not change the config on the switch. And yes, I did check it and even reloaded the config I had backed up just to make sure. It matches what it was previously and, the part that's driving me nuts, is 90% of the clients are working just fine. There's nothing I can connect to just these 10 users (they're running same OS --xpsp2-- as everyone else, are in various locations of the building (and hence connecting to the backbone at different points).

What am I missing??

The dhcp range given to the clients is 192.168.2.x, mask 255.255.0.0, gw 192.168.0.1.

Ok, that's an issue right up front. The client subnet and the gateway are in two different subnets. If you put them all in the same subnet (ex: 192.168.0.0/24) as the servers then the clients will be able to connect without issue.

If you plan on using separate subnets (only useful in a very large environment) then you'll need to do some routing to be able to cross subnets.

the part that's driving me nuts, is 90% of the clients are working just fine.

Open a command prompt window on the clients that are working and compare their IP information against those that aren't. Pay attention to the IP address. Are they in the same subnet? Are the ones that are working in the same subnet as the servers?

Sounds like routing was taking place to get those 10pc to the linux server and now that routing is gone.

Imagine the power of knowing how to internet search
http://www.lib.berkeley.edu/TeachingLib/Guides/Internet/FindInfo.html