DNS Resolution mechanism

[#1]

Hi guys,

I have configured one A Record with 3 different IPs,

www.company.com 10.10.10.1

www.company.com 10.10.20.1

www.company.com 10.10.30.1

I need to know what is the algorithm used by NIOS when a user is resolving www.company.com :

will it be  balanced ?

10.10.10.1     33%

10.10.20.1     33%

10.10.30.1     33%

is it Random ?

In My case DNS server returns the same IP 80% of the time, which is the first IP entered.

Thank You

[#2]

You can't configure one A record with 3 different IPs. Each A record has exactly one IP.

So you either configured:

a) 3 A records, each with the same name but with different IPs

or

b) one host record containing 3 IPs

However, for both a) and b) the default behaviour is round-robin, so distribution should be 1/3 for each of the three IPs.

Are you 100% sure there were no other queries to the server in between your queries ? The round-robin distribution is a total distribution across ALL queries received on the DNS server. It is not a per-client round-robin distribution.

So for proper testing/verification you'll have to ensure you are the only client sending queries to that DNS server at that time.

You can also setup different distribution/ordering schemes as per:

https://docs.infoblox.com/display/nios84/Enabling+the+Configuration+of+RRset+Orders

[#3]

Hi @evnull09 

Yes I have 3 A Records same name and different IPs.

The document talks only about host record "Enable setting RRset order for hosts with multiple addresses" 

Does it apply to simple A Records ?

Do I have to Add host with 3 IPs instead of 3 A Records with same name and multipe IPs ?

Thank You

[#4 · Accepted solution]

Hi,

that setting probably only applies to Host records. Not sure what the algorithm is if you setup 3 A records for the same label, like you did.

I did a quick test and 3 individial A records as well as a host record with three IPs showed the same behaviour. The answers were always round-robin so for 9 queries sent I got 3 replies with each of the three IPs on top of the list, so exactly the expected result of 1/3.

When you are testing, are you actually sending your test queries to the authoritative DNS server for the zone ? Or are you using a local cache/resolver ? If you are using a local cache then that cache might use another distribution scheme, so you should make sure to test against the auth server directly.