Wired Intelligent Edge

I'm seeing this wierd issue in our environment where the link between some MAS stacks will suddenly drop, lights out, no errors, no spanning tree warnings.

What could cause such a failure? Protocol or hardware?

Yann,

 To help understand better, can you elaborate more w.r.t below qns:

1) What model of switches (S3500/2500/1500) ?
2) Uplink ports are used as stack-ports or Front-panel (network ports) are configured as stack-ports ?
3) What version of code ?
4) This strange behaviour started recently? Just seen once or repeats ?

5) can you send the 'tar logs tech-support' ?

6) Dump the o/p of 'show stacking interface'

Plus, is stack split (show stacking members) ?

Thanks,
-Vinay

Here is some more information :

All switches are s3500 with the 10 GBE 4 port stacking module installed.

The ports that are going down are the first 2 ports of the module that are not configured as stack ports. (It does happen regardless of the switch/stacking module), These ports link to other stacks in the building.

We are running version 7.4.0.1

This behavior started a couple months ago but doesn't happen everyday, sometimes it will be several days before the issue comes back, other times it will happen overnight.

The stack appears complete when show stacking members is typed

The output of show stacking interface is attached as a txt file and while looking at it I noticed that the last link status change time lines up with the link that went down this morning...

What would cause a stack port link to change?

Attachment(s)

show stack inter.txt   8 KB 1 version

Yann,

I'm a little confused here. You stated "The ports that are going down are the first 2 ports of the module that are not configured as stack ports." but you're saying the stacking output is showing a link event at the same time (I see a few members with one link flap). We would need to see more detail in terms of up time, system logs, etc to get the bigger picture of the issue.

cappalli is right, I would open a TAC case regardless so they can look at the logs in greater detail to determine where the issue really is.

Best regards,

Madani

Madjali,

I'm as confused as you on that one... it's definitely a head scratcher.

I did open a case with TAC, I was just hoping for some insight.

Just noticed something on the port that is down...

GE4/1/1 is administratively Up, Link is Down, Line protocol is Down
Hardware is Gigabit Ethernet, Interface is N C, Address is 00:1a:1e:15:96:b3
Encapsulation ARPA, Loopback not set
Configured: duplex (Full), Speed (10 Gbps), FC (Off), Autoneg (On)
Auto negotiation in progress
Interface index: 706
MTU 1514 bytes
Link flaps: 1
Flags: Trunk, Trusted

Auto negotiation in progress

 This will be shown to any/all the oper 'down' copper interfaces.

But since these are fiber ports, and assuming some issue with auto-neg:

 - Can you dump o/p of

   show interface gigabitethernet 4/1/1 transceiver

       (Also for remote remote end port)

And enable below traceoptions

(MAS) (config) #traceoptions
(MAS) (traceoptions) #layer2-forwarding flags interface
(MAS) (traceoptions) #interface-manager flags all

After some time (when in problem state), collect 'tar logs tech-support' & attach it to the case for further analysis...

Thanks,

-Vinay

As a test, I have changed the transceivers on both ends of the link on 4/1/1 to 1 GB units, that link is currently stable (18 hours into the change)

Here's is the output from another link that gave me trouble that is currently working but still at 10GB.

(STACK-C) #show interface gigabitethernet 6/1/0 transceiver

Vendor Name                                : OPNEXT, INC.
Vendor Serial Number                       : L12E36915
Vendor Part Number                         : TRS2001EN-0065
Aruba Certified                            : YES
Cable Type                                 : 10GBASE-SR
Connector Type                             : LC
Wave Length                                : 850 nm

(STACK-A) #show interface gigabitethernet 0/1/1 transceiver

Vendor Name                                : OPNEXT, INC.
Vendor Serial Number                       : L12E38320
Vendor Part Number                         : TRS2001EN-0065
Aruba Certified                            : YES
Cable Type                                 : 10GBASE-SR
Connector Type                             : LC
Wave Length                                : 850 nm

I will enable the traceoptions on both ends of that link and report back.

Yann,

I should also capture the transciever details. You can add the keyword "detail" on the end of the "show interface gigabitethernet X/Y/Z transciever" which will pull stats from the SFP. It would be good to check if any levels are in alarm.

Best regards,

Madani