Comware

Hello,

I have 2 small buildings connected by Cat-6 from the computer lab building, and they have 2524's serving their offices. The lab uses an HP4000m fed by 1000FDX fiber from the Administration building.

I want to double up the uplinks to the small buildings with 2 CAT-6's to each switch, but I notice that the HP2524 only supports Trk1, not Trk1-Trk8 like the HP4000.

My main question is: Do the Trk numbers have to be the same on both ends, or can it be Trk1 to Trk1 for the small building to the lab's 4000m, and then Trk1(hp2524) to the lab's Trk2(4000m) ? I figure if they are of type trunk the switch only looks for that, not the actual Trk_ number..

Question #2:
Also, I have 2 HP4000's connected by trunking 4 ports for a maximum of 400 Mbps, but I was wondering if it would work to set up 2 trunks between the HP4000's and name them Trk1 for the 4 ports, and then Trk2 for the other 4 ports so I'll have 800 Mbps backbone between them or will this cause a loop in the network? (I wish I had money to order those cards with 1 RJ45 so I could have 1GB link from switch to switch)

*I read about the need to turn on Spanning Tree Protocol to do this, but I run an Apple network with LDAP (not appletalk), and a Novell Network with some IP/IPX, and I heard STP doesn't play well with those protocols.

Also, are you guys out there using ABC on your trunk ports, or just ports going to user machines and what did you set it to? I've set mine to 50 for the time being on backbones and 30 (default) for user machines.

Thanks for any help on the Trk numbering and on the maximum number of Trk links I can have from 4000m to 4000m.

-Walter

Hi Walter,

For your first question, as you suspected the trunk numbering isn't important. Trk1 on a 2524 will happily talk to Trk(X) on another switch.

Question 2, if you were to setup 2 trunks between 2 switches, it would create a loop unless spanning-tree was enabled. You really need the gigabit modules to get more bandwidth there unfortunately.

As you mention in your network, spanning-tree may cause some problems - this is usually because the 'fast' mode has not been set on the ports that the clients are connected to.

For ABC, I would just leave it at the defaults of 30 unless through experimentation you've found it to be better at 50%.

As a side note, I noticed in one of your previous postings that you were having an issue when telnetting to your switches it would time out and as a workaround you would ping the switch until it started responding and then telnet would work fine. I have a sneaking suspicion that ABC is causing this behaviour. I have also seen similar behaviour when running meshing on the 5300/3400 products.

Don't forget to assign points to posts that have helped you.

Matt

Matt,

Thanks for the quick reply the other day, we've been working on problems all week, and I have another question which you may be able to help me with.

We are using PRTG traffic graphing software which uses SNMP to our public community name on the porcurves to monitor building to building links, etc. I am plugged directly into an 8000M (10.3.1.1) and there is a 1 GIG fiber card going to a 4000M (10.3.1.45) in another building. That 4000m (10.3.1.45) running c_09_26.swi as of 2 hours ago is timing out 80% of the time on the SNMP requests, but another switch off of the 4000m (10.3.1.45), a 2524 (10.3.1.56) is always responding to it's SNMP when using this PRTG software. (My SNMP timeout is set to 10 seconds.)

I just can't believe that 1 hop away over 1 gig fiber and I'm timing out on stuff. Ping also doesnt always reply on the 10.3.1.45 switch and Telnet is also hit or miss.

My main question to HP or anyone: why does this happen? did I not set ABC correctly? is this a known issue, or is my network just really busy? What other settings can I change or tweak in protocol priority, ABC, or IP priority 0 - 7 ?

The other question is about the switch IP config menu in regards to the gateway. Our firewall to the internet is 10.1.1.1 so I make my switches gateway 10.1.1.1, but I read that it should be the IP of the switch 1 hop back from this switch, but it depends on if you are going to a different subnet. we are all 255.0.0.0 here, unless you're going to the internet to a different subnet. So should I make my switch gateways 10.1.1.1 (primary firewall) or make it the IP of the previous hop?

Lastly, ABC being set to 30.. what exactly qualifies as a broadcast packet? ARP? Ping? SNMP? and why limit them? do they create lots of unneccesary traffic or are they just a normal thing on a network and I should set my ABC back to 0 for all ports? Should I use ABC for uplinks to other switches or just to user machines?

Thanks for any help,

Walter

Hi Walter,

I'm assuming it was timing out before you updated the firmware..

Since the 2524 is responding fine, I suspect it's either ABC or Meshing that would cause this behaviour. I don't think it's Meshing though because I have run some tests with that in the past and I didn't see this happening. Don't want to completely rule it out though.

Since disabling ABC is easy, I'd recommend you try this and test. (If in a Mesh you need to disable it on all Mesh switches)

ABC by itself proxies ARP requests (which are broadcast packets) instead of flooding them out all ports. When enabling ABC it also sets the broadcast limit for each port to 30%. In theory with ABC enabled you can get away with having less routers and use larger sized subnets like you are currently doing. For most networks though I don't think you would really notice any real benefit by having it enabled. How many devices are actually on your network?

The broadcast limit can be enabled with or without ABC by going into the individual port settings. So if you disable ABC you may want to do this. The main advantage of enabling a broadcast-limit is to prevent a loop in the network completely bringing down the network due to a broadcast storm.

If I get a chance this week I'll see if I can test this ABC theory out myself.

Also I believe you should leave the gateway on your switches to 10.1.1.1

Matt