Stackwise Virtual Switchover time 70 secs??

[MATTHIAS SCHAERER]

I have a LAN core with SVL and generate traffic from and to a dual attached access switch. When reloading the switch in the SVL stack that carries the load it takes up to 70 seconds that it is balanced to the remaining switch in the stack. From the perspective of the access switch the PortChannel goes down immediately, so no reason not to switch to the backup link.

Is this like it is intended to be or do I misunderstand something about the feature?

The switches are C9500 (17.3.3) I have two SVL links that are up and running, do DAD over PAGP to two C9300 (17.3.3). 

[marce1000]

 - Check all logs , also on the 9300's , look for anomalies if any.  You may want to use a syslog-server and have all involved-devices use it, which will probably give a better overview to analyze logs.

 M.

[balaji.bandi]

Do you mean SVL Link connected Directly between Cat 9500 and DAD Link going via Cat 9300? Do you have high-level Physical topology and port configuration of both Cat 9500 and Cat 9300 ( 70 Seconds very high, some applications can not tolerate this)

[Leo Laohoo]

Raise a TAC Case.  

I have read about a bug with this sort of delay (but cannot remember the keyword).  Troubleshooting it is not an easy task.  It is best to engage TAC.  

[MATTHIAS SCHAERER]

To be more precise after some troubleshooting:

The LAN Layer2 part switches quite quickly, but if some L3 processes like upstream routing protocols are used they do not converge in an appropriate timeframe. It seems that there is no status information shared between the master and the standby CPU. Or that this at least seems to have room for improvement. We check with TAC.