I work for a web design/web hosting company that uses Spark Fibre for our office internet. We are on a business plan, not a residential plan.

As part of what we do we often update the DNS records on our clients domains in order to move them over to our hosting, I completely realise that DNS changes take time to propagate, that's not what the issue is.

What I'm noticing is that ns2.xtra.co.nz is usually pretty quick about picking up the updated record, but ns1.xtra.co.nz is much slower, and more alarmingly, for a reasonable period of time (a few hours at least) it will randomly return either the old or new IP address.

This is frustrating because it means that for a period of time we can only inconsistently access the new site, and I found it pretty concerning that a single server doesn't consistently return the same result, I can understand it being stuck on the old IP address but returning two different results within seconds of each other is very strange.

I've noticed this on a few different domains on separate occasions, so it doesn't appear to be a one-off thing.

I'm testing this using a command along the lines of "nslookup example.co.nz ns1.xtra.co.nz", so I believe I should be bypassing any caching on my end, whether in the OS or at the modem/router level.

I have attempted to contact Spark support using the usual channels but to be honest, the ones I've talked to don't seem to have the level of technical knowledge to understand what I'm talking about.

The other frustrating thing is that it will resolve itself after a (seemly random) period of time, so often by the time I notice the issue, and get through the hold times to manage to talk to someone beyond the first line of defense the DNS server has started being consistent again, even though it was definitely playing up all morning.

I do have another client domain that was updated today, which I suspect will be in the midst of the issue, however I can't confirm that it's playing up as I have a different ISP at home and can't query the xtra servers. I'm happy to PM this to one of the Spark guys for them to test.

So basically I'm asking, has anyone else noticed this issue or can confirm what I'm seeing?

Is there some secret code (shibboleet?) that would allow me to talk to someone at Spark who understands what I'm actually talking about, when the issue occurs again?

Thanks in advance,

James