On Thursday, 25 June 2020 18:29:03 UTC Paul Wouters wrote:
> On Thu, 25 Jun 2020, Mukund Sivaraman wrote:
> > For whoever is interested, this is a description of a pattern of queries
> > noticed at busy public resolvers that has led to issues in at least 4
> > different sites in the last 2 months.
> > 
> > The current revision is a work in progress. We are still developing some
> > mitigations for NIOS, and some more introductory text also has to be
> > added.
> 
> I would add a more explicit section on using prefetching of frequently
> asked queries, which mitigates (eliminates) the period when an answer is not
> available in the cache.

i'm sorry in advance for anybody who is now hearing this for the 100th time.

please don't pre-fetch. feel free to pre-discard; TTL is a maximum, and if you 
want to randomly subtract from TTL upon every reuse, you'll remain compliant. 
if the data turns out to be important it'll return. but prospective pre-fetch 
means you may be creating network load for no purpose -- the data may not come 
back again, or if it does, it may have changed, in which case you don't want 
the stale stuff that pre-fetch would have brought you.

this solves the thundering herd problem but without creating new problems. the 
goal is de-synchronization. random subtractions to TTL in cache accomplishes 
it.

-- 
Paul


_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to