[Pdns-users] SLAVE refresh performance

[Klaus Darilion]

Am 09.09.2011 19:40, schrieb bert hubert:
> On Fri, Sep 09, 2011 at 03:20:36PM +0200, Klaus Darilion wrote:
>> Currently ~10 000 zones are configured. Every slave-cycle-interval=60
>> PDNS checks if the "refresh" is expired, thus causing 10 000 SOA
>> queries, taking around 5 seconds.
> 
> Hi Klaus,
> 
> These are probably the 10000 queries on the local database?
> 
>> Is it just the way that PDNS queries all 2mio SOA records for refresh
>> detection? This seems a bit inefficient.
> 
> This may have to do with getting possible DNSSEC details for a zone. We
> never used to have to do this.

I just verified on a similar PDNS which acts as master (NATIVE). On this
I do not see these queries. Thus, I suspect the SOA queries are just for
SLAVE operations (calculating timers), not DNSSEC related.

> 
>> I currently use:
>>   distributor-threads=32
>>   receiver-threads=8
>>
>> Is there anything I can do to handle 2mio zones and still having a short
>> slave-cycle-interval?
> 
> Can you try 'query-cache-ttl=200' and see if this improves things? If it
> does, it would help us solve the issue.

Which DB-queries will be cached by PDNS? Only DNS-query related
DB-queries, or all DB-queries (e.g. slave-cycle-interval triggered queries)?

And what happens if the SLAVE updates the DB due to an incoming AXFR.
Will it also expire relevant cache entries?

Thanks
Klaus