Hi,
I have the following DC fabricpath scenario.
- two Nexus 6001's as vPC spine running L3 with SVI's with HSRP version 1
- two Nexus 6001 as vPC leaf
I want to migrate HSRP version 1 to version 2 on the SVI's because of ipv6, using the following approach:
- remove HSRP config from SVI that is in HSRP standby -> this SVI is no longer participating in HSRP group
- remove HSRP version 1 config from remaing Active HSRP group on SVI (ie. the other Nexus), and quickly re-apply Version 2 config
This will cause the HSRP group to reinitialize and become active again after a while, thus causing a short disruption.
Tweaking the HSRP timers to msec values helps to minimize the disruption from about 10 seconds to a much more acceptable value. (I configure the timers to their default values just after the HSRP group reaches the active state, Nexus in vPC doesn't like non-default HSRP timers.). After that i reconfigure the SVI on the first device to HSRP version 2 so it will participate in the group again..
When simulating this on a stand alone Nexus 6001, i noticed that exactly when the HSRP group becomes Active, a Gratuitous ARP is sent that will update the ARP cache of the connected machines. This actually works fine.
Now i tried this in the datacenter vPC spine/leaf setup, mentioned above, and to my surprise there is no Gratuitous ARP anymore when the HSRP group becomes active. I could not find any info about this.
Does anyone have a clue why these GARP's are not there? Or perhaps a better way to change HSRP version from 1 to 2 with minimal disruption?
I noticed that despite of the absense of GARP, pinging the HSRP virtual address does actually renew the cached arp entry fast enough, Perhaps GARP is not necessary?
Thanks,
Erik