Cisco Catalyst 4507RE Chassis Reboot in VSS Mode During Active Sup Failure

[shunmubala]

Hi All

I’m not sure if any of you have encountered such issues but my company recently purchased 2 units of Cisco Catalyst 4507RE that comes with quad Supervisor module 8L-E. We enabled VSS in our staging environment, and this is good.

However, I noticed when I unseat/remove the active supervisor module in Core Switch #1, in order to simulate hardware failure on the active supervisor module, I observe the following behavior;

The secondary supervisor module in Core Switch #1 doesn’t become active, instead the secondary supervisor module in Core Switch #2 becomes active.

The entire Core Switch #1 chassis reboots. It takes about 20Min for the entire Core Switch #1 to come up.

This means all LAN switches and Servers connected to Core Switch #2, doesn’t face any interruptions, which is good. However, those LAN switches and Servers connected to Core Switch #1, those with single leg/connection, NOT DUAL/REDUNDANCY, will be down for almost 20Min.

My vendor opened a Cisco TAC case 682352147, and was informed by Cisco TAC that this behavior is normal/expected on the Cisco Catalyst 4507RE. In my opinion, I find this a bit odd, as I never noticed such behavior with Cisco Catalyst 6509E with quad Supervisor modules or just 2 supervisor modules (one on each chassis) running VSS, in my past working experience.

Hence, my company has decided to enable HSRP instead. Perhaps, VSS in Cisco Catalyst 4507RE has matured yet, as compared to Cisco Catalyst 6509E.

Can anyone shed some light on this please, if you’ve Cisco Catalyst 4507RE that comes with either quad Supervisor module 8L-E or just 2 supervisor modules (one on each chassis) running VSS.

Thank you.

[Philip D'Ath]

I haven't tried running quad supervisors.  I know you do need quite recent software to get good support.  What software version are you running?

I assume your VSS link connects all supervisors?

[shunmubala]

Hi,

With regards to my newly purchased Catalyst 4500E Supervisor Engine 8L-E : Catalyst 4507R+E Switch, the IOS XE version that I'm running on is 3.8.3E(ED). 

I was told by Cisco TAC that even if I were to upgrade to the latest firmware, the product behavior remains the same i.e. in the event I were to unseat/remove the active supervisor module in Core Switch #1, in order to simulate hardware failure on the active supervisor module, the entire Core Switch #1 chassis reboots. 

Cisco TAC case 682352147

[nilapemanohar.pradhan]

When ever you have Dual Sup on a single chassis, this behavior is bound to happen. Since the active sup is holding the VSS configuration with the Secondary switch. When ever the Primary switch's  Secondary sup is building the configuration from the Active sup...The Secondary Switch has become active on VSS.  Reason behind, the Secondary switch sees the Primary switch is switch is down. Even after the Primary switch has the secondary sup built the configuration is still waiting for Active and Standby election. Either you remove the secondary SUP and both the primary and Secondary switch. So the fail over scenario is faster and you dont have major outage.

[shunmubala]

Hi Sir I disagree with your explanation because in C6509E, when the active Sup fails, nothing gets rebooted, and everything is seamless. It's only with C4507RE+ that the chassis reboots :-(  Please kindly advice if I've misunderstood your explanation.

[Reza Sharifi]

Hi,

That is the correct behavior when you are running quad sup VSS regardless of the platform. It is so much easier and simpler to just use one sup in each chassis and not 2.

HTH 

[shunmubala]

Thanks Reza. Please do correct me if I'm wrong but if that's the correct behavior, why on earth is Cisco selling VSS solution in C4507RE. I think it's just silly. I mean you don't get this silly behavior in C6509E chassis. When the active sup fails, the entire chassis doesn't go into reboot mode.

I know if I had dual leg for all my LAN switches and servers, I wouldn't be complaining about my current situation, but sadly this is my situation, and I have to deal with single leg situation :-(

[Reza Sharifi]

Hi,

The way VSS is designed is that all access switches, servers, etc.. must be connected to both chassis, if not VSS is not going to be very useful. If you have access switches, server that are not connected to both chassis than you really don't have redundancy.

HTH 

[shunmubala]

Yes sir, I am fully aware that dual leg is always the best option. I agree with you. However, my question here is why in C4507RE+ the chassis has to reboot when active sup fails in vss mode, but you don't find such behavior with C6509E?

[Leo Laohoo]

but you don't find such behavior with C6509E?

The two models, 4500 & 6500, may have Cisco on it but both are "owned" by different business units (inside Cisco) and they have different development teams.  

[Leo Laohoo]

I know if I had dual leg for all my LAN switches and servers

If the servers are critical then they should, at least, have dual upstream links/legs to the switch.  Having a critical server with a single upstream link is just asking for trouble. 

why on earth is Cisco selling VSS solution in C4507RE

The way VSS works is totally different to a stack switch system like the 3750-, 2960S/X/SR- an, 3650/3850 series family of switches.  

With VSS, there has to be a way to stop a client from sending data down a link to a dead chassis.  The only way to force a "switchover" of the traffic is to reboot the chassis.  

[Reza Sharifi]

Hi,

Not sure what version of software you are running but it seems that there is an update in ver 3.8.x regarding quad sup VSS.

If the VSS Active switch or supervisor engine fails, the VSS initiates a stateful switchover (SSO) and the former VSS Standby supervisor engine assumes the VSS Active role. The failed switch performs recovery action by reloading the supervisor engine.

In Quad-Supervisor VSS mode, if the VSS Active switch or supervisor engine fails, the VSS initiates a stateful switchover (SSO) and the former VSS Standby supervisor engine assumes the VSS Active role. The in-chassis standby (ICS) on the failed switch becomes the VSS Standby and the former VSS Active becomes the ICS for the VSS Standby.

If the VSS Standby switch or supervisor engine fails, no switchover is required. The failed switch performs recovery action by reloading the supervisor engine.

http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst4500/XE3-8-0E/15-24E/configuration/guide/xe-380-configuration/vss.html#85440

HTH