Proliant ML530 trunking to 4000M Switch

[Scott Smith_15]

We're using a Compaq ML530 server and trunking both ports on a Compaq NC3134 NIC to a 4000M switch. The problem is, we are having problems getting it to communicate with the rest of the LAN. We need to increase the server's outgoing bandwidth, so I set the adapter to Switch Assisted Load Balancing. The 4000M switch ports have been set to "Trunked", "FEC", and not trunked at all. Depending on what combination of these settings I use, NT Explorer on the server only sees a limited, varying number of the computers on the LAN (this includes the other servers too). Even when a particular server is displayed, Explorer will not open some of them - I get a "Network Path not found error". Some do open though. A bit of background; we're running both TCP/IP and IPX on our servers, and varying combinations on clients. The documentation on the SLB Teaming configuration states it should handle both protocols without problems, but changing the bind order of the protocols on the server also alters which servers/clients it can see. To eliminate the NIC, I've also tried this with Intel Pro 100+ server adapters, with similar results. All firmware/software involved is the most recent available (except the switch firmware, which is C.08.22). The switch configuration is nearly "as received" from HP - no VLAN's or other segmenting has been done. Can I get this to work, or is there a better way? Any help would greatly appreciated.

[Gonzo Granello]

Scott,

looks like the way you have set up the original config is as good as it will get with the NIC's you have. The Switch assisted Loadbalancing would need a switch that supports that, the 4000M wont. In my opinion FEC at the switch and server (which is the hardware protocol so to speak) would be the choice plus some kind of load balancing on the server side as I know the teaming SW provides should do. If you need mainly outgoing bandwith, the switch is not involved in the loadbalancing as it would be the other way around. However, Trunking for that matter is not bandwith aggregation per se! What do i mean? Here is a quick example: you have two 100M links trunked, hence 200M, well sort of. If the amount of traffic let's say on NIC1 reaches 150M with that backup you are doing overnight, you will drop packets and have only 100M traffic since a TCP conversation is pending on the MAC address where it started and there is no known trunking algoryhtm that can split up traffic to use two connections. I hope that makes sense... To answer your question, more bandwith would be achieved with up to 4 100M links (assuming you have lets say two) or upgrading to gigabit ethernet between the server and switch (hoping that the server Hardware can keep up). The 4000M also has newer Firmware you can get from the web, Gigabit Fiber and Copper are also (still) available, the update of Firmware should be becoming routine for all equipment as it usually makes life easier in the long run.

Again, hope that helps and points are apprechiated.

GG

[Scott Smith_15]

Gonzo,

Thought that the 4000M supports FEC, which, according to the info I've seen, will aggregate the ports and pass whatever protocol I'm running. The teaming driver on the Intel adapters specifically states they support FEC (both static and dynamic), and I've set it to static with the results I noted. Though I'm hardly an expert, the switch documenation says it'll work with FEC - is that wrong? Today, I tried something really complicated - I tried to ping several other items on the same subnet- and got nothing but timeouts. The teamed "adapter" will ping itself, but nothing else. That's the real problem. BTW, buying a gigabit interface for the switch and NIC's for the servers isn't an option - no funds. Segmenting the LAN is possible, but would create a lot of other headaches. Since supposedly, it will work, I'd like to use the teaming option if at all possible. I just can't get the blasted thing to communicate.

[Gonzo Granello]

Scott,

FEC on the 4000 will definitley work, I used that in several configuratons. Just make sure that all the ports you want to use are configured as Trunk-members and the trunk type is set to FEC on all of them. Once you have done doing this step, you want to make sure that the trunk is configured to be in the same VLAN as the other nodes you want to communicate with. Not sure if that applies to your config, but by default the trunk will be in the default vlan once configured - even if all ports where in another VLAN before. Since the switch can't know what vlan you need it will revert to the default VLAN. That could be one reason why you can only ping yourself and no one else. Hope that helps, if not we could get in touch more direct and i'm sure i can help working it out.

GG

[Scott Smith_15]

I've tried every combination of every setting there is on every piece of hardware. Still won't communicate when trunked with SOME network hosts. There is no difference in configuration between the ones it does see, and ones it doesn't that I'm aware of. I'm thinking its an ARP problem of some sort, because of what's happening. I tried another experiment - I set up LMHOSTS to ID the other servers (with WINS still set on the server I'm trying to use). Same problem when both cables trunked, but when one cable was removed, and the switch still set to FEC, the computer can browse normally. VLANS are not enabled on the switch - do I need to turn them on? If you're willing to give it a shot, please email me at SmithSC@nnsy.navy.mil, and let me know what other info you need to shed some light on this. Thanks for the assist.

[Scott Smith_15]

Problem turned out to be an intermittent network cable. Reterminating the cable corrected the proble.