[Pdns-users] Help needed debugging knot_control timeouts

[Chris Hofstaedtler | Deduktiva]

Hello Klaus,

* Klaus Darilion via Pdns-users <pdns-users at mailman.powerdns.com> [221213 08:40]:
> We regularly use "pdns_control retrieve" (via TCP) to trigger zone transfers. Very often we receive:
> Timeout error: Error from remote in receive(): Resource temporarily unavailable
> 
> Our PDNS is quite busy, plenty of NOTIFYs, SLAVE checks and incoming AXFRs.
> 
> What exactly happens when "pdns_control retrieve" is called? Is the "retrieve" asynchron (zone gets added to AXFR queue) or is it synchronous (client + server hast o wait until PDNS has time to do the transfer)?

It is async; addSuckRequest is called and that writes into the
d_suckdomains queue. It has to obtain the communicator d_data lock
however.

> For us it would be good even if PDNS is busy, the "retrieve" request should be answered instantly and the zone just added to the AXFR queue. Or does adding the zone to the queue also requires some locks that cause the timeout?

I wonder if its contending with the periodic slaveRefresh
(slave-cycle-interval/xfr-cycle-interval).

Do you see a "long" running "info-all-slaves-query" query at the
same time?

Chris

-- 
Chris Hofstaedtler / Deduktiva GmbH (FN 418592 b, HG Wien)
www.deduktiva.com / +43 1 353 1707