Server Dropping Clients 3

[dougp23]

We have just upgraded from Catalyst Express 500s to 3560s. I have two servers that are just giving me fits! In one case, the client PCs and server are on the same switch! Users will be in the application, sometimes not doing anything, and messages start popping up "Unable to access file MAIL.DAT in F:\DATA\USERS." And we must close the app and restart.

Almost always, all users go down at the same point. What is so strange is my port on the 3560 shows no errors....and the server shows no errors. I have tried setting the NIC and switchport for speed and duplex, I have tried issuing "switchport mode access". Sometimes it will be great for a few days, then it falls apart again.

One server is a Compaq/HP, and one is a Dell.

Ideas anyone???

 

[enzo288]

Create a seperate VLAN just for servers. Also configure each switchport for spanning-tree portfast. This may help.
The NIC on the servers and the switchports will be configured for auto-detect so checking speed & duplex is not the answer. Also check the statistics on the individual switchports which the servers connect to.

CCNA, CCDA, CCNP, CCDP, almost CCIE

 

[dougp23]

Hey enzo.

Yup, the servers and PCs are in Vlan1, the phones are in Vlan5. I just added spanning-tree portfast for the servers, as well as "switchport mode access".

Crazily enough, the counters on the ports that have the servers usually show very little, if any, errors. One shows 8 CRCs.

 

[enzo288]

VLAN1 should not be used. Create a VLAN for the users, one for the servers, and one for switch management. Also, depending on the size of your network you might need 1000 Mbps connection to the servers. If you have a large amount of users hitting these servers over a 100 Mbps connection, it may be causing a bottleneck.

CCNA, CCDA, CCNP, CCDP, almost CCIE

 

[dougp23]

I am just curious as to why VLAN1 should not be used? The phones have thier own Vlan (Vlan5) and swithc management has its own Vlan (Vlan9). Also, my busiest server has 8 users....

One of my servers has a Gig card, and when this trouble started, the first thing Dell said was to cut that speed down to 100 Full. And it was great, for a week, then it went lousy again.

I appreciate your help! Why not Vlan1 though???

 

[voltron1011]

Servers should be on their own Vlan for many reasons. Firstly, they are chatty and are broadcasting stuff left and right. They should be in their own broadcast domain so that they aren't eating up bandwidth and flooding other hosts.
Secondly, security to those servers is questionable. Any user can walk up and plug in a Linksys Hub and loop the network without even realizing it. This will cause a Spanning tree storm that would bring your network to its knees.

 

[baddos]

VLAN1 shouldn't be used because it's the default administrative VLAN. You have changed that to a different VLAN, so disrguard that comment.

As far as backing your server down to 100/Full, you need to make sure that the switch's settings match that.

 

[dougp23]

Thanks everyone! We put each building in a separate VLAN, it appears one building was doing a "broadcast storm". Now, we can dd seelct individuals to other VLANs if they need it, and the really clueless users, we can just leave in some "nowhere" VLAN.