Forum Discussion

Blimmer_282204's avatar
Blimmer_282204
Icon for Nimbostratus rankNimbostratus
Dec 04, 2018

stop advertising virtual address when pool members down

Hi all,

 

I have 2 virtual servers running on the same IP with different ports (443 and 2000) and therefore sharing the same virtual address (lets say 10.1.1.1 for example). This setup is repeated on the backup site (same IP, VS's etc.). Now the team looking after the nodes need to be able to pull down the nodes on one site in order to upgrade, thereby failing traffic over to the backup site. To do this we want to stop advertising the virtual address on the main site when the nodes for the https VS go down, but NOT bring the virtual address down completely, just stop advertising the route. I believe we can do this with iCall but am unsure where to start - any help appreciated!!

 

application delivery
BIG-IP
iCall

2 Replies

Sort By
  • jurgenvdmark_14's avatar
    jurgenvdmark_14
    Icon for Nimbostratus rankNimbostratus
    Dec 04, 2018

    I'[m not exactly sure how this works when you've got 2 sites with same IP-addresses. What you can do is install an advanced routing license on the BIGIP's and use a dynamic routing protocol like OSPF to advertise your virtual addresses. You can advertise the same addresses with different costs

     

    Every virtual address is then advertised as a /32 for IPv4 or /128 for IPv6. When all VIP's sharing the same virtual addresses go down the virtual address is no longer advertised to OSPF

     

    Hope this helps

     

  • daboochmeister's avatar
    daboochmeister
    Icon for Cirrus rankCirrus
    Dec 04, 2018

    If i'm understanding correctly, you have two different IPs, one at each site ... if that's correct, and if all clients use hostnames to access the VIP, then the usual way to manage this would be DNS. If you don't have a global server load balancing (GSLB, like is provided for by e.g. F5 Big-P DNS Controllers), then you would simply change the resolution in your DNS system to the backup site's IP, and back to your primary after. Note that if the DNS records involved don't have a short TTL, you'll want to shorten the TTL ahead of time (e.g. to a minute or whatever), so clients don't wait long to get the new IPs at failover/failback. You can set the TTL back after - but probably don't want to, since if you really did have to failover to the backup site, you'd want the resolution to change quickly anyway, to meet whatever SLAs you have.

     

    If the IPs are the same, like described in the above reply, or if they're different but clients use the IPs directly to reach the VIPs, that won't work. May want to detail out your situation more fully if so.