SMLT

[GerhardBade]

Hi folks,
we are running a LAN with 12 Passports 8606/8610. Used features are OSPF, VRRP and SMLT.
We have a strange problem with our SMLT and hopefully one of you can help.

First the network structure (our Passports are named NBBxxx):
- NBB50-1 and NBB11-1 are our Core
- NBB11-1, NBB50-2 and NBB50-3 are the Serverfarmswitches
- NBB01-1, NBB12-1, NBB32-1, NBb34-1 and NBB72-1 are Backboneswitches to connect the building to the LAN, and Accessswitches are connected to them.

- The Backboneswitches are connected over Layer3/OSPF to the core
- The Serverfarmswitches are in one IP subnet and connected over Layer 2 with MLT to the core
- The Coreswitches are building the default gateway over VRRP for the SMLT connected Serverfarmswitches.

We have also a few Vlans, with spread over the whole campus.
These are in one separate Spanningtree group on all Passports.

Normally, everything works fine, but sometimes we have this problem:
If one passport fails in case of power fail, or fiber link break, some users can not reach all servers.
Example: Backboneswicth NBB32-1 fails.
User 1-10, connected at NBB72-1 can reach only server 1,2,4,5.
User 11-20, connected at NBB72-1 can reach only server 3,4,5.
Same for all other users. We checked the forwarding database at the Serverfarmswitches and the two Coreswitches. All looks good. The only solution is to boot the two Coreswitches. We have this since we implemented SLMT at the Coreswitches.

But booting the Coreswitch if one Building-Backboneswitch fails is not acceptable.
What causes this forwarding problem and is there a table at the Coreswitches we can flush to solve this?

Any idea to help us?

Kind regards.

 

[NMSman]

What version of software are you running on these switches ?

 

[GerhardBade]

3.7.17 on all, with 256MB RAM extension per CPU

 

[GerhardBade]

We are working with 8690SF-CPUs in the Core.
In the Backbone we have a few swicthes with 8690SF CPUs and a few with 8692SF CPUs.
In the Serverfarm with run 8691SF and 8692SF CPUs.

But allways same CPU-Type per Switch

 

[NMSman]

Gerhard,

I have seen similar problems with newer software (4.1), where multicasts and sometimes broadcasts are going missing across MLT/SMLT trunks when a failure occurs somewhere else in the network. This is in a very similar setup to the one you have (OSPF/SMLT/VRRP).
At the moment we are looking at replacing hardware to see if the problem is hardware related.

 

[GerhardBade]

I didn´t expect that replacing hardware will help.
We interchanged CPUs from Backboneswitches and Coreswitches. It did not help.

 

[NMSman]

We are looking at a possible problem with the 8608SX cards. We seem to have problems where there are 8608SX cards installed, but not with the E cards