When running FGCP is there any way to maintain a separate MGT IP on the active and passive FGTs?
I can see that the data-plane interfaces on the units need layer two reachability because in the event of a failover the IP and MAC will float to the standby unit (and GARP will take place). It seems a bit awkward to have the same MGT IP float between the boxes because you wouldn't get any direct SSH/SNMP/HTTPS reachability of the standby for monitoring purposes.
Additionally, if one used dynamic routing on the FGT, wouldn't you need to peer with the virtual IP on the upstream switch? Again, this seems a bit awkward compared with FGSP.
Yes I had a similar issue where the policy install didn't succeed when using the dedicated management interface with FortiManager.
As mentionned by the KB article, it is recommanded to use another interface to communicate with Fortimanager so that you don't loose access to the cluster in case of a failover. The dedicated management interface is intended for SNMP monitoring and direct access to the secondary device.
I'm just updating this thread with some closing remarks in case it helps others.
Basically there are two features which would be interesting to allow independent management of the units when running FGCP a-p mode.
#1: ha-reserved-management/ha-direct/ha-mgmt-interfaces. There are too many touchpoints and workarounds with this approach and they're not well documented. I had to fix local-in, then FMG and after all that I couldn't get outgoing DNS requests to work.
#2: Use a standalone-management vdom as documented here:
The Fortinet Security Fabric brings together the concepts of convergence and consolidation to provide comprehensive cybersecurity protection for all users, devices, and applications and across all network edges.