Swift failover in Fabric

[JP10]

Hello Cisco community,

Recently we've tested a failover of one server connected to the Fabric. The server has 2 management interfaces, each connected to a different edge switch, one interface is active and the second one in standby mode. When the fabric edge or link of the active interface goes down, failover automatically switches standby to active mode. 

We discovered, that time between the whole process takes around 2 seconds, which is for a real-time server quite much. In the old network, we didn't have this issue (running only on the L2 layer). Is there a possible configuration in DNA-Center or Fabric, which could shorten this time, ideally to instant switch (I don't know, secondary lisp entry for the same address or so)?

Thank you in advance for your suggestion!

[jalejand]

It depends on how fast the second interface comes up. And with "going up"  I mean how long it takes for the interface to send an ARP (usually GARP) so the device is registered in device-tracking and then part of a dynamic EID on LISP to create the CP registration.

Are the management interfaces using the same IP always? If they use the same IP, then this is a LISP mobility case of use.
If they use different IP addresses, then you can use static device-tracking entries to have a placeholder for the EID.

Additionally, how did you notice that the communication is down for 2 seconds? It is a monitoring tool? Inside or outside of the fabric?

[JP10]

Hi Jalejand,

Thank you for your reply. The management has unfortunately the same IP address and we are obligated to use the same IP. We have a control server inside the fabric, in the same VN.

[jedolphi]

1-2 seconds is standard latency at the moment for a wired roam on fabric. There is some things in  fabric that take time such as MAC and IP theft detection. There might be some enhancements in future to accommodate this scenario but there's not much you can do about it today. Cheers. Jerome

[andy!doesnt!like!uucp]

@Jerom

sounds sadly...

[jedolphi]

Hi Andy, everyone, we are working on a solution to this. Please reach out to your sales team or CX team to discuss roadmap and time frames for accelerated wired roaming. Cheers, Jerome