Forum Discussion

SSHSSH_97332's avatar
SSHSSH_97332
Icon for Nimbostratus rankNimbostratus
Jan 04, 2012

Specifying unicast addresses for failover

When specifying unicast, rather than multicast, addresses for failover, you

 

need to specify two pairs of IP addresses, where one of the pairs indicates

 

the management IP addresses of the two units. The other pair indicates two

 

self IP addresses, one for VLAN HA on each unit.

 

 

 

Why is that , if the idea is just to let each unit sense the other , why to use management Ips , it was sufficient to use Vlan HA IPs ?

 

 

 

i mean :

 

bigip_B|192.168.15.17|192.168.15.18|1026

 

bigip_B|10.10.10.2|10.10.10.3|1026

 

 

 

 

one of them ( the HA vlan ) was sufficient , why to put management Ips ?

 

 

 

8 Replies

  • mgmt interface/ip is always available while tmm interface/selfip is not. for instance, while booting bigip, tmm might not be ready fast enough and ha pair might end up with active-active if only selfip is configured in unicast pair since unit is not able to reach peer's selfip.
  • and for this to be useful , should the management interfaces sense each others ( be connected at same switch for example )

     

    also what if i did not put the entry for mgt , will HA work fine
  • and for this to be useful , should the management interfaces sense each others ( be connected at same switch for example ) it is not required to connect them to the same switch.

     

     

    also what if i did not put the entry for mgt , will HA work fineyes, it may work even not having mgmt ip in unicast pair but it might not work in some condition too. so, the best practice is to configure it.
  • Thx nitass , but the customer still need any sitiuation at which it may not work without mgmt , can u give me any good example

     

  • can u give me any good example for example, in ha pair, if we reboot unit1, unit2 will be active. in case if redundancy state preference is none, unit1 won't preempt an active role when coming back. anyway, if sod process is ready before tmm, sod may force unit1 to be active since unit1 is not able to communicate to unit2 via tmm interface/selfip.
  • pr's avatar
    pr
    Icon for Nimbostratus rankNimbostratus

    Hi Nitass,

     

    So does it means that if self IPs goes down and mgmt ip is till remians up then failover b/w the devices wouldn't occur ???