Automatically control downlinks according to uplink's state

[TaoLeiCCIE_2]

Hello.

A switch has one uplink and several downlinks.

Is there any mechanism to accomplish this demand:

  When uplink gets down, all downlinks get down automatically.

  When uplink recovers then all downlinks recover, too.

Thanks.

[andrew.prince]

Is this a Layer 2 or Layer 3 requirement?

[hobbe · Accepted solution]

Hi

There are some things you can do.

First of all if the switch supports EEM then you can use that to setup a script on what will happen when a link goes down aswell as when it goes up.

There are several EEM scripts that will help you with that already or is easily modified to do just that.

This i have done in the past, works very nice.

If you need help with this just let us know.

IP-SLA should be able to do thesame but i have no personal experience with it.

Depending on how you do things in your network you can get systems to react on different things like Logging or SNMP

You can setup logging/traps so that the logging/snmp server (if it is not over the shutdown link ofcourse) will login to the switch and shutdown apropriate interfaces.

Just some examples on how to do it, there are most likely more.

Good luck

HTH

[TaoLeiCCIE_2]

Hi, all.

It is a layer 2 requirement.

Actually I'm finding the equivalent of H3C's "Monitor Link" in Cisco products.

The "Monitor Link Overview" part of the following page describes the technology:

http://www.h3c.com/portal/Products___Solutions/Products/Switches/H3C_S5810_Series_Switches/White_Paper/200912/656299_57_0.htm#_Toc215567385

Thanks.

[hobbe]

Without reading more then just the few first sentences of your link i would state that this is not what you asked for in your initial statement.

This is STP ie if one link goes down another will go up and if one is up the other will go down to protect the network from loops forming.

What you asked for was a link that if uplink goes down, then the other links will also go down and if uplink comes back up then the other links will follow and also go up.

ie the totally opposite.of STP

.

[andrew.prince]

why do you want this?

Sent from Cisco Technical Support iPad App

[TaoLeiCCIE_2]

I think I indeed have to describe our scene:

|      |

A -- B

|      |

D    C

\    /

server

A, B, C, D are layer 2 switches. The server has two NICs, one is primary, the other is secondary.

By the support of server software, configuration of the two NICs (IP, MASK, GW, MAC, ...) can be identical.

At any time only one NIC is enabled (primary), the other is in standby mode (secondary).

When primary link is down, the server software detects this event and immediately enables the secondary NIC to continue transffering tranffic. When recovered, the server disables the secondary NIC and uses the primary one again.

We do *not* use STP.

Suppose the primary path is "server -> D -> A".

If link between A and D gets down, the server cannot detect this, it continues transfering traffic on its primary link to D.

Therefore, for switch D (and switch C), a mechanism is needed to make the downlinks to down state when the uplink gets down, and vice versa.

Who has read the web page in my last post will notice that "Monitor Link" technology is just what we need.

The reason I created this discussion here is I wonder whether there is a Cisco's solution for this situation.

[andrew.prince]

Cisco and the IEEE have a solution to this, it's "spanning tree"

Sent from Cisco Technical Support iPad App

[TaoLeiCCIE_2]

If STP could be the best solution anywhere, then there is no need for Cisco to invent "Flex Links" technology.

In my opinion, STP is no use here.

If you have read my desciption of our scene carefully and keep the topology unchanged, can you explain how STP achieves our goal, please?

[andrew.prince · Accepted solution]

Firstly Flex Links are "not" a replacement for Spanningtree - they are an "alternative" to spanningtree for people that do not want to run spanningtree

Secondly Flex Links do not provide loop avoidance

Thirdly you have quite a few limitations when using flex links - like they can only be in 1 pair, they cannot be in an etherchannel, port security could trigger false positives and close a good link down......

So the real question is - if the Flex Link technology was not created, what would the solution be.....spanningtree.

To your question "If you have read my desciption of our scene carefully and have seen the topology carefully, do you explain how STP achieves our goal, please"  Very Simple

Current Topology

|      |

A -- B

|      |

D    C

\    /

server

New Topology

|      |

A -- B

|  X  |

D    C

\    /

server

You cable D to B and C to A, with spanning tree the link between D to B would be blocking, as D has a direct link to A (Assuming A and B are primary and second root bridges - which would be the best topology) and the link between C to B would also be blocking.

If the D2A link does down, spanning tree automatically puts the link between D2B into forwarding and the same goes for the C2B link if the C2A goes down.  If the server primary nic 1 goes down - there is not change in spanningtree as the server is not creating a loop.

[TaoLeiCCIE_2]

So, I summarize the following points:

1. As you said "...they are an "alternative" to spanningtree for people that do not want to run spanningtree", yes, we indeed do not want to use STP if conditions permit.

2. If use STP, we have to change our topology (add two links).

Thank everybody for your kind replies.

[andrew.prince]

I really think that you need to read the Flex Link config guide - as - looking at the topology you have provided it will not work for you. Using what you stated as the requirement

"If link between A and D gets down, the server cannot detect this, it continues transfering traffic on its primary link to D.

Therefore, for switch D (and switch C), a mechanism is needed to make the downlinks to down state when the uplink gets down, and vice versa."

Using your topology

|      |

A -- B

|      |

D    C

\    /

server

To configure the Flex Links feature, you configure one Layer 2 interface as the standby link for the link that you want to be primary. With Flex Links configured for a pair of interfaces, only one of the interfaces is in the linkup state and is forwarding traffic. If the primary link shuts down, the standby link starts forwarding traffic. When the inactive link comes back up, it goes into standby mode.

So in your topology where is your backup?? you only have 1 link between D and A, if the link between them does down, where is your other "Flex Link" ?? To use Flex Links - you are going to have to add more cable connections......

[TaoLeiCCIE_2]

Hi.

You're right, "Flex Link" does not fit. I mentioned about "Flex Link" in earlier posts and my purpose was to give an example that sometimes we might need an extra mechanism rather than STP.

I find that the "Monitor Link" feature of H3C's switch is a good solution for our situation. So I wonder whether Cisco has an equivalent feature. (It seems not)

BTW, both "Flex Link" and "Monitor Link", their reconvergence speed are very fast - millisecond level, which is a big advantage over STP/RSTP/MSTP.

[andrew.prince]

OK - so you are really dead set on using the H3C technology - crack on, Cisco does not have a solution that 100% does the same(that I can find).  So you are going to have to rip out your existing switches and purchase new ones to replace them....good luck with that, seems a waste to do that for 1 feature when you can use existing equipment, and it already kinda does it already.

I have just 1 question - why are you so against using spanningtree, and running 2 more cables?

[TaoLeiCCIE_2]

Hi.

From the point of designing, we can add more cables and use STP/RSTP, this is no problem.

But the reconvergence time of STP/RSTP is at second level, not millisecond level, thus it does not fit our demand.

The server is transffering realtime multicast IPTV traffic, so we expect a very short failover time.

Fortunately, we currently have both cisco and h3c devices in hand.