3110x stack in IBM bladecenter chassis

[Jack Leung]

Hi all

I have 4x 3110x switches in a Bladecenter H chassis and the stack cables are connected so that they are in a circle (IE, SW1 connects to SW2, SW2 connects to SW3, SW3 to SW4 and SW4 to SW1). Each of the switches physical 10GE port is trunked to a 6509.  If I pull a 3110x switch out, does this network connectivity? Theoretically it should not?

TK2EWRBC4SW01#sh switch neigh

  Switch #    Port 1       Port 2

  --------    ------       ------

      1         4             2

      2         1             3

      3         2             4

      4         3             1

[Reza Sharifi]

Hi,

If you pull one of the 3110X out, it should not effect your network, but I would test it in the lab first before doing it in production, so you know the behavior.  3110x series are stacked just like the 3750 series, except the stacking cable is different for the 3110X series.  Also, make sure to assign priority to each switch, so you always know which switch switch is the master. 

HTH

[Jack Leung]

I just tried it and it definitely does break connectivity. When the switch was removed and replaced with the new I did get a version mismatch status which I expected. However I then started to see sluggishness and eventually lost all access to the stack.  Pinging the management interface resulted in some successful and some not. I'm confused. What is the process to upgrade this if the new switch is running a different IOS and obviously the 3110x cant be booted up on its own unlike a 3750 which can be booted on its own and have the IOS upgraded before introducing back into stack?

[Krzysztof Zaleski]

Hello,

I would like to follow up because I am about to face the same issue. We need to upgrade the stack, but want to make sure that one of the two switches is properly switching data (of course, short time for datapath switchover is expected). Is it possible or the longer downtime is most likely expected?

Best regards,

Krzysztof

[Dirk Woellhaf]

Hi everyone,

Updateing a new Member in the stack can cause some high cpu usage which can cause some ICMP-Packets not being  answered. The Update-Process has also a really low prio on the stack-ring itself. Thats the reason why updating a member does take quite a while. It should NOT affect the traffic-forwarding to your Blade-Servers.

How did you realize the uplink to the Cat.6509? Classical STP or Etherchannel? I have seen some issues on SVI when a STP-Uplink was going down.

regards,

    Dirk

(Please rate if helpful)