Mystery Network Problem

So I'm tossing this out there to see if anyone might have some thoughts on where to look because I just find this baffling.

We had a Cisco 3750 Catalyst switch set up with a number of hosts connected to it. All of these hosts were configured to do NIC teaming with etherchannels set up on the switch. We then added a second 3750 to create a stack. We re-configured the etherchannels on the existing servers so that each one had one connection to each switch. We then added a number of new servers with each one using etherchannels (one connection to each switch). Everything seemed fine. All of the servers were available on the network, etc. But, we noticed that of the 12 new servers, two of them had this odd issue. All of the servers are in the same VLAN and IP's in the same subnet. All of them can get out to the internet and can be reached by *most* computers in the same VLAN and in other VLANS. BUT...the two oddballs can't talk to the other servers connected to the same 3750.

I turned off the NIC teaming and disabled the secondary NIC port. We removed the switch ports from the etherchannel so it was just a normal, plain jane port. Still had the problem. Moving the server IP information from the primary to the secondary NIC port cleared up the problem...at first. While trying various permutations of the NIC teaming (trying to track down the problem) things got the point where if the server was configured to use NIC1, it could get out to the internet and other various servers, but couldn't talk (no ping, no thing..they don't even see ARP requests) to the other servers on the switch (EXCEPT those in a different VLAN) or some other random machines on that same VLAN.). Setting the server to use NIC2 resulted in it being able to talk to servers it previously couldn't talk to, but it suddenly couldn't talk to the gateway and so couldn't reach anything else. It was previously able to do so.

What this FEELS like is that the switch has somehow declared these two server's MAC addresses as persona non-grata and won't route traffic from or to it except to seemingly random exceptions.

Anyone have any thoughts on where to look as I'm pretty much stumped at this point.

Reply to
Aaron
Loading thread data ...

The 3750 only has a limited amount of TCAM that must be used for routing (CEF entries) switching (for MAC entries), multicast and qos queues. Sometimes the "default" allocation of TCAM isn't right and you have to adjust it.

Do the MAC entries for the servers exist in "show mac-address-table", if not you probably have an TCAM issue. Also do a "show mac-address-table count" and see how many entries are left for MAC addresses (this will be the last line of the output). If you are doing layer 3 on the switch then you probably don't have enough TCAM allocated for CEF entries, and the reason you are having issues.

The "show sdm prefer" command will show you what is currently configured. The default is "default" but we have had to change this on 3750's that we are doing layer 3 switching on. To change it, issue a "sdm prefer " command, and you will need to reboot for it to take effect.

Reply to
Thrill5

3750.http://www.enterprisenetworkingplanet.com/nethub/article.php/3527301>
Reply to
Aaron

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.