9800 to 10K Migration

[capitanGiaco]

Hi all

we just moved from the omniswitch 9800 to the 10k (OS10K 8.3.1.314.R01 GA, September 07, 2016).
In the first week from the installation we noticed slowness in the remote networks.
After one week remote switches begin to block themselves.

every switch is configured with SPT 1x1:

show spantree mode
Spanning Tree Global Parameters
Current Running Mode : Per VLAN,
Current Protocol : N/A (Per VLAN),
Path Cost Mode : AUTO,
Auto Vlan Containment : N/A
Cisco PVST+ mode : Disabled
Vlan Consistency check: Disabled

no one ports was blocked by the SPT.

the only abnormal thing that I saw is the cpu usage in the slot 1 of the 10k, the cpu hit the 100%!
Than I disconnected each port from the slot1 until I found and isolated the remote switch that caused the malfunction.
I repeated the disconnect operation in the remote switch Omniswitch 6250 (6.6.4.177.R01 GA, May 24, 2013) and find a crushed copper cable.

How is possible that a fault cable connected to a 100mbit interface on a remote switch can block all my network ?
(Probabily the cable was already crushed at the time of the 9800)
Is there in the 10k (with AOS8) a feature to prevent this kind of troubles ?

thanks

Giacomo

[silvio]

did you see error frames at
> show interfaces counters errors
or
> show interfaces 1/1

Link-monitoring is a feature to shutdown a port if there are a lot of errors:
> interfaces 1/1-24 link-monitoring admin-status enable

Have you seen a lot of stp topology changes (show spantree vlan xx)?
regards
Silvio

[capitanGiaco]

Hi Silvio

No Error Frames, just some little lost frames where the interfaces is on a LACP.
Again nothing of strange with show show spantree vlan xx

Topology Changes : 1,
Topology age : 7 days and 05:30:43,

Giacomo

[silvio]

Hi,
you have written "After one week remote switches begin to block themselves."
What does it mean?
- remote switch is access switch (direct attached to core)? or remote ....
- what is block? uplink are blocked? permanent blocked? any loops/redundand ways?
- remote switches are also omniswitches?
show spanntree ports blocked

Did all switches know about there root (10k) in every vlan? (show spantree xx xx=vlan)

regards
Silvio

[capitanGiaco]

Hi,
you have written "After one week remote switches begin to block themselves."
What does it mean?
- remote switch is access switch (direct attached to core)? or remote ....
- what is block? uplink are blocked? permanent blocked? any loops/redundand ways?
- remote switches are also omniswitches?
show spanntree ports blocked

Did all switches know about there root (10k) in every vlan? (show spantree xx xx=vlan)

regards
Silvio

Hi and thanks for the answer!

the switch is access/distribution (direct attach to the core and clients are connected to it).
No ports were blocked, just the cpu to the 100%
When show health all cpu show 99-100% network traffic die...
The 10k is the root for each vlan (8192), but it wasn't a spanning tree problem, there weren't blocked port!
Just the cpu of the remote switch and of the slot 1 (where is the remote switch connected) to the 100% of utilization.
Changed the crushed (short circuited) cable everithing is working fine.

As asked in the first post: The only thing that I don't understand is why the high utilisation of the cpu was replicated to the slot 1 of the 10K, and why after this the 10K isn't able to manage (packets drop) network traffic in every ports (although other slot cpu was to 10-15% of utilisation).

Giacomo