Forum Discussion

mshoaib's avatar
mshoaib
Icon for Altocumulus rankAltocumulus
Sep 18, 2024

Adding New BigIP Units Using Free CPUs on i7800 (LB13 & LB14) to an existing HA Pair

Hello,

We have a pair of i7800 devices, each running two guests with the following configuration:

i7800 Unit 1: LB09 & LB11
i7800 Unit 2: LB10 & LB12

 

Current Setup:

LB09 & LB10: These are part of a High Availability (HA) pair, running in an Active/Active configuration.
LB11 & LB12: Similarly, these two guests are also in an HA pair, running in Active/Active mode.
Each LBx unit is utilizing 6 CPUs (which is the maximum allowed per guest).

 




The i7800 comes with 14 CPUs total, so there are 2 unused CPUs per unit (not allocated to any guest).

 

Questions:

  • Can we bring up new BigIP units (LB13 & LB14) using the 2 free CPUs on each i7800 unit?
  • Can these new units join the existing HA pair of LB11 & LB12 and sync configuration to each other?
  • Can the new Units serve their own traffic groups ( as shown below in the diagram ) ?
  • Is there any potential limitation or issue with utilizing the remaining CPUs for this purpose?
  • Will i7800 run out of memory or go low in memory ?

 

Goal:

  • Offload 20% of the traffic (a handful of VIPs) to the newly created LB13 & LB14 units.
  • All LBxx ( 11,12,13 & 14 ) should back each other up and sync configuration across

     

    Thank you for any assistance 

 

active active
application delivery
high availability
redundancy
  • Sebastiansierra's avatar
    Sebastiansierra
    Icon for MVP rankMVP
    Sep 23, 2024

    Hi mshoaib,

     

    About your questions:

     

    • Can we bring up new BigIP units (LB13 & LB14) using the 2 free CPUs on each i7800 unit? Yes it´s possible, this link shows you the Possible vCPU allocation per guest per slot/appliance  https://my.f5.com/manage/s/article/K14218 
    •  
    • Can these new units join the existing HA pair of LB11 & LB12 and sync configuration to each other? Yes, it´s possible, but in the case of the fail of the nodes LB11 & LB12 with 6 CPUs, is possible new LB13 & LB14 in the traffic group can´t support the load with only 2 CPU.
    • Can the new Units serve their own traffic groups ( as shown below in the diagram ) ? Yes, you can select the preferred device to run the traffic, and the failover order in case something fails.
    • Is there any potential limitation or issue with utilizing the remaining CPUs for this purpose? No, the VCMP is a virtualization platform so you can use all the resources available, take in mind that the system has its exclusive CPUs to run the F5-VCMP system.
    • Will i7800 run out of memory or go low in memory ?No, the same answer as the last one for this topic.

     

    About your goals.

    • Offload 20% of the traffic (a handful of VIPs) to the newly created LB13 & LB14 units. In my opinion is better if you deploy the new devices without introducing them to the existing traffic group, and then you can migrate the applications to these new devices.
    • All LBxx ( 11,12,13 & 14 ) should back each other up and sync configuration across. As I mentioned, in the case of a failure in nodes 11,12, the load couldn´t be supported for nodes 13 and 14.

    Hope it works.