How to reduce downtime when migrate to an AE interface

[AllanGao]

Hi All,

Am going to bundle an existing layer3 interface (e1/1)with extra one (e1/2 ) to an ae1 interface. And then move the ip address from e1/1 to ae1. 

This is in a HA A/P configure, question is how to reduce the downtime to roughly 0?

If it will impact existing sessions?

As zone name not changed but ingress/egress interface, and PA sessions matching tuple not related to interface, but still has a concern for any outage.

[S.Cantwell]

Hello

The configuration to move the IP to AE is not too bad.  What you may not be considering is that the AE needs to be added to the routing table as well, so there are additional changes that need to be implemented. 

It may not be feasible to suggest how to make downtime equal or close to 0. 
It is most reasonable to expect to have about 5 min of downtime during this period, as you need to include commit time and for sessions to be established, as you already have noted. 

What other questions can we answer for you?

[AllanGao]

Thanks for the reply, Steve. 

Yes, I forgot to mention the VR routing table part, , will change it to ae1. 

1. So which means it will impact the existing sessions and which need to be reestablished?

2. If things like that, just wait for the new session setup or need to do 'clear session all' ?

3. Actually confusing part is if Fast path matching the existing session will notice the ingress/egress interfaces change? As zone is not changing.

Thanks in advance

[S.Cantwell · Accepted solution]

HI @AllanGao 

Yes, you can expect that there will be some time (maybe 30 to 60 secs) for the sessions to be re-established on the dataplane. 

You could (do a clear session all) and let all of the sessions get started up again.

Fast Path will be matching zone (not changing), but the interfaces do change, so Fast Path will be affected.  This is why a clear session all is not that bad of an idea, to let the FW establish Fast Path again.