Spanning Tree Questions

[sbyrd]

I might be having a spanning tree issue with a corporate customer. He has 2 Cisco 6500e on each end of a link. Between his link I have an AF24 bridge powered by WS-Minis at each end.

Whenever he connects his Cisco directly to the Netonix he claims he gets a Spanning Tree block and his traffic will not pass through the Netonix. If he connects a small HP-1700-8 procurve switch between the Netonix and the Cisco he can pass traffic.

According to the HP manual for the 1700-8 - "The ProCurve Switch 1700 supports pass through (flooding) of spanning tree BPDUs so that upstream devices on the network which support Spanning Tree Protocol can detect loops that exist on the ProCurve Switch 1700 interface."

Now on the WS-MINI I have spanning tree disabled on all the ports. Is that the correct setting? Will the WS-MINI pass spanning tree properly from his Cisco's with it off, or should I have spanning tree on?

I will add that I have never used STP in my network and really am quite unfamiliar with how it operates outside of a very basic understanding of its function.

Also when he connects the Cisco directly to the Netonix the CPU usage on the Mini goes to 96%. When the HP is between the 2 switches the CPU usage is normal.

[sirhc]

Our switch does not pass BPDU packets, a most other manageable switches also will/should not pass BPDU packets.

Try turning RSTP on on the WS-6-MINI switches (both sides) but I thought you had it ON first and then I had you turn it OFF as well?

If you never had it ON try it ON but the answer is we are a managed switch and thus do not pass BPDU packets.

[sirhc]

Also I am not sure "why" this company would have RSTP even enabled on a port that is plugging into a switch that only has a single path to the other building?

Personally I would tell them to disable RSTP on the port that plugs into the WS-6-MINI.

But if they want RSTP enabled on a single port that only has ONE path to the other side then enable RSTP on the WS-6-MINI

But I still say something else is going on here meaning they have something configured wrong in their network.

Obviously there is some type of weird BPDU or other special packet storm occurring from their network to our switch which is what is causing the CPU to spike in the WS-6.

[sbyrd]

Thanks for the input. I have presented this to them and we are going to hopefully troubleshoot some more soon. While they only have one link now between buildings the have asked us to quote them on putting a fiber path in as well. That would mean later they would need STP as they would have the primary fiber link and the backup AF link.

[sirhc]

Well if they want RSTP then you need to enable RSTP on the WS-6-MINI

Also if you arrange remote access Eric will look at the WS-6-MINI

Not sure why they would want a FIBER and an airFIBER.

I understand the idea of backup but if you're hanging fiber you hand 2 pair fiber and drop the wireless.

I could see adding a backup wireless link but the reliability of the AF24 units is like 99.999% uptime

[sbyrd]

Well if they want RSTP then you need to enable RSTP on the WS-6-MINI

Also if you arrange remote access Eric will look at the WS-6-MINI

Not sure why they would want a FIBER and an airFIBER.

I understand the idea of backup but if you're hanging fiber you hand 2 pair fiber and drop the wireless.

I could see adding a backup wireless link but the reliability of the AF24 units is like 99.999% uptime

The customer is a major book publisher and they want for every building 2 separate paths. They have 2 buildings in the Business Park already and this one is their 3rd. I was hoping the AF would be sufficient and I would not have to go to the trouble of putting in fiber, but it is what they want. At the end of the day it just allows me to charge them more per month as the fiber path has to run across our fiber so he is paying me lease fees.

I have talked to him and we are going to either try it with RTSP or off. Once we have RTSP on and configured or off, I will check the CPU problem. If it continues I will most definitely get with Eric to look at it remotely.

[sbyrd]

Sirhc - Today at 3:30pm EST we are going to test with RSTP on. However, I have been told by the customer that if it does not work then I have to remove the Netonix switches to see if the problem is resolved. If removing the switches resolves the problem then they must stay removed. He says that this will be justified by the fact that the Netonix switches are keeping the bridge from working as promised and intended.

I do not want that to be the result. Can you and or Eric be available by phone and setup to access my system remotely at that time should the test with RTSP on fails?

I can email or PM you the information necessary to access the equipment.

[sirhc]

A managed switch is not a transparent bridge.

If they want to use RSTP between these buildings then you will need to enable RSTP on our switches.

A managed switch will not pass BPDU packets.

BPDU packets are what is used for managed switches to talk and exchange information.

[sbyrd]

A managed switch is NOT a transparent bridge.

If they want to use RSTP between these buildings then you will need to enable RSTP on our switches as well and they will become part of the RSTP.

A managed switch will NOT pass BPDU packets.

BPDU packets are what is used for managed switches to talk and exchange information.

I know that now as I was definitely wrong in my understanding of how the Netonix switches functioned (we are a pure routed network). However, as I mentioned earlier when their switch is connected to the Netonix (with STP off) the CPU on the Netonix is pegged. You indicated earlier in this thread that you believe they are possibly doing something wrong to cause this.

So what I am saying is if I turn on RSTP and they connect directly to the Netonix and not only does it not work, but the CPU is pegged are you or Eric available to remotely access the switch and see what could be going on? Can you provide me 'backup' as it were to tell the customer that the Netonix switches are NOT the problem and his config is the problem?

[sirhc]

I will ping Eric and see if he can contact you to arrange access to the switches to try and figure it out.

This may not be easy to tell from afar what they are doing wrong.