TTLs on A records?

[Barry Margolin]

In article <gtd8nt$1vdr$1 at sf1.isc.org>,
 "online-reg" <online-reg at enigmedia.com> wrote:

> Hi All:  I'm running Bind 9.5.0-P1 / Fedora on my primary NS.
> 
> Are TTLs on individual A records universally supported?

They're supposed to be.  Many DNS-based load balancing systems and 
services depend on it.

> 
> I have a domain with a TTL of "3h", and I wanted to route traffic between 
> two servers in that domain quickly, so I set the TTL to the A record like:
> 
> www    300    A    123.123.123.123
> ;www    300    A    123.123.123.124
> 
> so I could uncomment one and comment the other to manually switch between 
> them.
> 
> I've had that setup for several weeks during testing...and I just reversed 
> the records, incremented the serial, and reloaded BIND.
> 
> On my secondary NS (Bind 9.5.0-P1 / Freebsd 7), when I dig the "www" record, 
> I see the TTL counting down from 300 (Cool!), and after it reaches "0", the 
> IP address resets to the new one....perfect!

A slave server is authoritative, not caching, so it shouldn't count down 
the TTL at all.  Or did you mean something else when you said "secondary 
NS"?

> 
> On my Windows DC (server2008), the change was also picked up after 5 
> minutes.
> 
> When I use some other lookup services, however (like samspade.org), the old 
> IP address shows up for much longer...like it's caching it and ignoring the 
> TTL for the record. Should I expect that behavior?

No.  Maybe the web site itself is caching.  Try querying your ISP's DNS.

-- 
Barry Margolin, barmar at alum.mit.edu
Arlington, MA
*** PLEASE don't copy me on replies, I'll read them in the group ***