2

I would like to know to which member does the request will be first directed to when NONE preference is set in the BIG-IP load balancers.

From this link, there are three ways to set the redundancy state preference. Now, if we doesn't specify the active unit and put the preference to simply NONE, then which unit will take up the request first?

Also, if the request is taken by some unit first, then what if the unit fail which took up the request fails in such a way that it could not transfer the control to another unit?

user12458
  • 123
  • 5

1 Answers1

2

If the Redundancy State Preference setting is "None", then neither cluster member will forcibly take over as the active unit. That is, whichever unit happens to be active will remain active. Alternatively stated, whichever unit completed the boot process first would be active (since it saw that its partner was down), and the second unit to finish booting should respect that decision.

If both units fail, then… well, you have a problem!

200_success
  • 4,771
  • 1
  • 25
  • 42
  • Is the user's request directly received to the load balancer first? – user12458 Dec 11 '14 at 07:10
  • Whichever load balancer boots up and tries to become active first is the active one (assuming none are up). That load balancer will be the one replying as that's what active means. Standby is waiting for the Active to have a problem so it can take over. It has nothing whatsoever to do with a 'request flow'. As for the rest of your questions about the specifics of the failover process.. you should be asking your sales engineer those questions. –  Dec 11 '14 at 07:15
  • I mean when the user makes a request from his web browser, after the DNS lookup, will the request be **first** transferred to the load balancer? – user12458 Dec 11 '14 at 07:15
  • That makes no sense. First of *what*? Either a load balancer is up and can accept the request or there isn't. 'transferred' is not a proper term for whatever you're trying to ask. It's about as descriptive as *cloud enabled* is. –  Dec 11 '14 at 07:16
  • The Virtual IP that the client connects to is hosted on the active unit. The active unit informs its router of that fact via ARP. So, you could say that the client's request makes some number of hops through the Internet, and the last router forwards the packet to the active unit because that's what its ARP table says. – 200_success Dec 11 '14 at 07:20
  • Ok. Also, what if the active load balancer which takes the request first, is failed for some reason after taking the request and is unable to direct it to the other load balancer? I mean what if the communication fails between the load balancers? – user12458 Dec 11 '14 at 07:22
  • 2
    Then it would act as any other failed connection. I don't usually configure my F5s for connection mirroring. The added overhead isn't worth the price of someone having to hit reload or reconnect their app due to a failover. Failovers shouldn't be happening regularly. As for communication *between* the load balancers.. they'll stop exchanging configuration information but failover will still work if you're using hard-wired failover. If you're using network failover and both load balancers are NOT connected to the same switch.. then you should've used the hard-wired failover. –  Dec 11 '14 at 07:28