Forum Discussion

Felkor_29336's avatar
Felkor_29336
Icon for Nimbostratus rankNimbostratus
Dec 17, 2013

Score bonus & failover condition

Hello guys,

 

A question regarding the behavior of the failover in the 11.4.1

 

Two devices, in active/standby, with a trunk

 

Active unit bonus score : +10

 

Bonus per available trunk interface : +10

 

Initial setup (F5A active):

 

F5A : 20 (2 int) + 10 (active) = 30

 

F5B : 20 (2 int) + 0 = 20

 

With 1 interface shut on F5A :

 

F5A : 10 (1 int) + 10 (active) = 20

 

F5B : 20 (2 int) + 0 = 20

 

=> F5A still active, ok.

 

With 2 interfaces shut on F5A :

 

F5A : 0 (0 int) + 10 (active) = 10

 

F5B : 20 (2 int) + 0 = 20

 

=> failover on F5B.

 

With 2 interfaces OK on F5A :

 

F5A : 20 (2 int) + 0 = 20

 

F5B : 20 (2 int) + 10 (active) = 30

 

=> F5B still active, ok.

 

Until now, everything is logic.

 

With 1 interface shut on F5B

 

F5A : 20 (2 int) + 0 = 20

 

F5B : 10 (1 int) + 10 (active) = 20

 

=> Failover on F5A !

 

I don't understand why we have a failover event in this last case. Any explanation ?

 

availability
fail over
management
preempt
score
  • afedden_1985's avatar
    afedden_1985
    Icon for Cirrus rankCirrus
    Dec 17, 2013

    We use a trunk with 2 interfaces and this is what we ended up doing. We don't need the 2nd 10 gig interface active so we only have 1 interface up per trunk and this is working ok.

     

    The weight we assigned was 60 and we have 2 interfaces so each interface has to be up to get a weight of 60, with 1 interfaces down the weight is 30 and on the active unit you add the active bonus making the total 55 on the primary as shown here and the highest score becomes active.

     

    LTM1:Eval:Active:In Sync] config tmsh show sys ha-group HA1 Sys::HA Group: HA1

    State enabled Active Bonus 25 Score 55

     

    On the standby it looks like this

     

    (LTM2)(cfg-sync In Sync)(Eval:Standby)(/Common)(tmos) show sys ha-group all Sys::HA Group: HA1

    State enabled Active Bonus 25 Score 30

     

    Consider when you activate the 2nd interface in the trunk. Now the full weight of 60 is added to the active bonus of 25 so that would be a score of 85 and the standby would be 60. If you lose 1 of the 2 trunk interfaces on the active the score changes to 25 + 30 or 55 which is lower than the standbys 60 so a single interface failure out of 2 in this configuration would trigger a failover. Consider how you want this to work fail over from 1 interface or both? So when we do activate the 2nd 10g interface we could increase the bonus to 35 so the one interface weight of 30 + active bonus 35 (65) would still be higher than the standby which is at 60.

     

  • Felkor_29336's avatar
    Felkor_29336
    Icon for Nimbostratus rankNimbostratus
    Dec 17, 2013

    Hi,

     

    Thanks for your answer. This is indeed a way to avoid this situation.

     

    But here I'm trying to understand why we have this behavior (failover or not in case of equality of the score). It looks like it is link to the management IP for the decision...

     

    • uni's avatar
      uni
      Icon for Altostratus rankAltostratus
      Dec 17, 2013
      Yes, it uses the management IP to break a tie
    • Ryannnnnnnnn's avatar
      Ryannnnnnnnn
      Icon for Altocumulus rankAltocumulus
      Feb 20, 2014
      Sorry to bump an old thread. I'm curious to know how the management IP is used as a "tie breaker"? is there a SOL that details this?
  • arpydays's avatar
    arpydays
    Icon for Nimbostratus rankNimbostratus
    Mar 08, 2016

    Dylan, it is in the link posted by nitass, here is the extract.

     

    About matching device utilization values In rare cases, the BIG-IP system might calculate that two or more devices in a device group have the same lowest device utilization score. In this case, the BIG-IP system needs an additional method for choosing the next-active device for an active traffic group.

     

    The way that the BIG-IP system chooses the next-active device when device utilization scores match is by determining the management IP address of each matching device and then calculating a score based on the highest management IP address of those devices.

     

    For example, if Bigip_A has an IP address of 192.168.20.11 and Bigip_B has an IP address of 192.168.20.12, and their utilization scores match, the BIG-IP system calculates a score based on the address 192.168.20.12.