Forum Discussion

Narendren_S's avatar
Narendren_S
Icon for Nimbostratus rankNimbostratus
Sep 22, 2017

BigIP_DNS VS Capacity Method

Documentation of VS Capacity Method in AskF5 says;

 

The Virtual Server Capacity load balancing method distributes DNS name resolution requests to pool members (virtual servers) based on a system-generated list of pool members (virtual servers) weighted by capacity. BIG-IP DNS selects the pool member with the greatest capacity most often, but over time, all pool members are returned in responses. When pool members have the same capacity, BIG-IP DNS uses the Round Robin method to select a pool member.

 

https://support.f5.com/kb/en-us/products/big-ip-dns/manuals/product/bigip-dns-concepts-12-0-0/2.htmlunique_1262380958

 

I have a query here. What are all the parameters logically decides the capacity of a pool member/virtual server?

 

If the project requirement is to distribute DNS based on the_LTM_Active_Pool_MemberCount_per_DC, will it be a good decision to go with VS Capacity Algorithm ? or is there any better algorithm to meet this requirement?

 

2 Replies

  • romar's avatar
    romar
    Icon for Nimbostratus rankNimbostratus

    It does not seem to work for me this VS Capacity method, at least in 11.6. In the logs can't see any trace of how a particular VS was selected. Somewhere in a devcentral article saw that "VS Capacity" metric is "how many pool members" for a VS. Here is my test: (1) defined on an LTM 2 VS, one with 4 members, the other with 2 members. The LTM was added to GTM as Big-IP server with "VS discovery" option. Then I have defined on the GTM a pool with the 2 VS from the LTM and made it single member to a Wide IP. The pool was set with preferred "VS Capacity" LB. Defined a listener w/ GSLB enabled. Then I am "dig"-ing the WIP hoping to see only the "4-member" VS. To make sure DNS caching does not get in the way I have set the TTL to 1sec. Tough luck though - I see both VS are more or less randomly selected - it's not a round robin, can get the 4-member VS 3 times in a row then the 2-member 4 times in a row etc - overall is 50%-50%. Enabled debug-level GTM logging - logging to an external syslog server - can't see any trace of the VS choice decision. To add to this - using iqdump can't see any specific VS-level metrics coming from the LTM. Not sure how this is supposed to work w/o GTM polling the LTM for metrics.