Forum Discussion
NimbostratusHigh-Availability over Management Interface
So, the situation is that the serverfarm 1 was down for quite sometime during OS upgrade and the failover did not occur for the vCMP guest on F5 - A on the Standby Unit F5 - B, but instead all the vCMP guests started working as Active/Active. Not to prepare the setup resilient, testing and config needs to be done.
Considering the diagram attached, here are the scenarios for testing the failover conditions:-
- Link 1 is down - Data link is down - F5 can be configured for failover using vlan safe and gateway safe
- Link 2 is down - Mangement interface goes down - The keep-alive info will not get exchanged and hence both the devices will start acting in Active/Active mode
- Link 3 is down - The result should be same as that of test 2 ??
- Link 4 is down - The device sync group will break but should not result into outage. HA vlan (non-routable) has been configured.
- Link 5 is down - The result should be same as that of Test 2.
I have not been able to see any document on F5 website, where it shows that the F5 failover can be controlled using the management IP address.
Simon_BlakelyEmployee
As detailed in
K14135: Defining network resources for BIG-IP HA features (11.x - 15.x)
adding the Management IP as an additional HA Unicast address is definitely recommended practice, to avoid a single network failure causing an Active/Active situation.
- Simon_Blakely
Just re-reading your diagram - the management interfaces need to be directly routable to each other, so that HA status is directly communicated between the two LTMs over the management network.
If link 3 goes down, HA status between LTMs is still communicated via HA network (link 4) - Active/Standby will still work correctly.