SGE2010P ring of 4 switches reports one "link down"

[ConniesPizza]

Any help appreciated. I called Cisco Support but they didn't have an answer.

We have 4 SGE2010P switches in a ring using ports 24 and 48 to connect. All are same hardware and firmware and the stack is working nicely.

Problem is - System - System Management - Stack Management page of web interface shows one link in the ring down yet it does not appear to be down. If we pull a cable from one of the other links then this "down" link is magically back.

Two questions:

1. Is the switch purposely breaking the chain to avoid a loop condition? If so, does that mean a 1-2-3-4 ring requires switch 1 to go the long path through 2 and 3 to communicate with 4?

2. Even if all is working as a ring and this is just a status reporting error, we have a problem in that one link failure cannot be detected since it shows down even when it isn't. That is bad for remote management and I cannot see any way to determine status of all 4 links shorts of visiting all switches and physically eyeing the ports in question.

Any insights appreciated.

Thanks.

John,

[David Hornstein]

Hi John

I believe the answer to question 1 is yes, as I think the switch uses a  least cost algorithm to as you said "break the chain" in a stack.

I would have thought John, that if you remotely bought up the web interface  you would see either three or four switches in a stack. Obviously three switches showing in the status screen indicates a failure.

regards Dave

[ConniesPizza]

Thanks, Dave.

Certainly I can see a ring that has been broken twice resulting in two chains. Goal, however, is to see first break of ring which results in a chain so I can fix it before a second break creates two chains and thus destroys the stack.

As it stands now, I cannot tell if a "link down" message is a real physical failure causing ring to be a chain, or just a logical adjustment by the switch to avoid a loop. Only after another second physical failure would it become clear that the first was physical and not logical and that is too late to do me any good.

I also see a fault in the least cost algorithm as a ring provides option for data to travel shortest direction. If ring is purposely broken into chain by the switch then some paths become longer.

In short, I would hope ring would remain instact and internal routing (STP for example) would handle shortest path and avoiding looping communications. However most important is for me to be able to see a physical failure remotely at first occurence.

Any ideas how to do this?

Thanks.

John,

[David Hornstein]

Hi John

With this stack, you haven't had to worry about spanning tree, the stack takes care of making sure that a broadcast storm isn't created by a loop, one of the interfaces becomes inactive. But if you had a real interface failure or a stacking port was to fail, the ring topology would still allow all switches in the stack  to still  be connected.

Well stacking becomes convenient for a few situations;

• reduce the number of infrastructure IP hosts that need managing on your network
• Allowing for Link Aggregation from servers to multiple switches in a stack.

You always have the option for;

1. taking the switches out of stack mode
2. running Link Aggregation with RSTP as per the rough diagram below;

The better more proactive solution  is to use a ONPLUS appliance, which will proactively inform you when there is a problem on your end user network. 

ONPLUS is more than just a cloud based service, there are alot of benefits of this appliance in fascilitating a pro-active support for your end user.

We produced p/n ON100-K9 appliance  to allow our partners to provide a more proactive professional services enviironment for their customers .  Even though my personal network can stand a failure, I utilize a ON100-K9 unit for demonstration purposes .

Because I use my network for experiments and it may change daily, the onplus appliance (p/n ON100-K9) send me quite often emails warning me of changes via the cisco cloud based service.  Here is an example below of a warning i receive via email.

For an independent description of ONPLUS appliance and the cloud based service,  check out this review from smallnetbuilder

http://www.smallnetbuilder.com/lanwan/lanwan-reviews/31805-cisco-onplus-reviewed

John , I am happy to also have a quick chat via phone  to discuss the topology options and better understand your needs.  If you wish send me your contact details to;

dhornste  at  cisco.com   (just remove the spaces and replace the at with @  )

regards Dave

Systems Engineer

Cisco Systems, Inc.

Research Triangle Park

NORTH CAROLINA

[ConniesPizza]

Thanks for all these details, Dave.

We want stack mode so we can manage all switches with one page and so we can use LAG on multiple switches.

My concern now isn't that the stack isn't a ring, but that I can't see whether it is or not. In proper ring operation it shows one link down. That misinformation hides whether that link is actually physically down or just displaying that way. Is there any way around that?

The illustration you have with LAG ring on the switches is intended as a substitute for stack mode? We actually need a LAG ring anyway just to keep transfer rates between the switches decent.

Thanks.

John,

[Peter Marquis]

Hi john,

     It looks like its Spanning Tree Protocol(STP) bringing down the link, I'd make sure the switch knows they are stack ports, I'm not 100% on how the SGE's stacking works, but that seems to be the look of things, check the STP status on the ports either side of the "down" link, i bet one of them is in blocking mode

[Tom Watts]

Hi John, I'm not sure if I read this right but a "LAG RING"? Essentially you went a "LAG" over the stack ports to get a 2g link between the switches?

This won't work. You will need a SX500 series for the 5g/10g stack ports if you'd like this functionality.

-Tom
Please rate helpful posts