> On Sep 13, 2018, at 9:39 PM, Alex <mysqlstud...@gmail.com> wrote:
>
> No, I should have mentioned that it doesn't always fail, and mailspike
> (in this case) doesn't have a quota. It also fails for my registered
> barracuda and other RBLs we pay for.
>
> It failed about a thousand times over the course of the day, while at
> least tens of thousands of queries with the same config have
> succeeded.
Some nameservers have rate limits and drop packets when you go too fast.
In my DANE survey I was seeing dropped queries and long delays with a
couple of hosting providers, with total throughput down to ~1200 qps
(which really meant ~240 domains per second as the survey does multiple
queries for many domains). I asked the two providers to raise the
limits, which they did, and now I'm doing ~2700 qps or 540 domains a
second.
In any case, this is really NOT a Postfix question, your local
resolver is returning SERVFAIL, debug your resolver, not Postfix.
For high query rates, perhaps the resolver needs more cache RAM,
a larger pool of open sockets for outgoing queries, ... Or, more
likely, the problem is upstream.
--
Viktor.
