[Pdns-users] Race condition during AXFR while updating slaves with two supermasters in high availability

[Pieter Lexis]

Hi Guy,

On 9/17/19 8:17 PM, Guy Lee wrote:
> I'm running in to a race condition problem with two supermasters in a
> high availability setup.
>
> Zone example.org has both Master1 and Master2 defined as masters in the
> database. I make an update on Master1, and it sends a NOTIFY to Master2,
> Slave1, and Slave2.
>
> Sometimes the slaves request the AXFR from Master2 instead of Master1
> who sent the NOTIFY. Except that Master2 hasn't updated itself yet and
> doesn’t have the updated serial, so there's nothing to update on the
slaves.
>
> I worked around it by setting the SOA RETRY value to 1 minute, so it
> checks for a newer serial quickly after the update, just in case it was
> missed.

This is a bug, there is a PR to fix this[1], but this PR has to be
un-conflicted and re-reviewed.

Cheers,

Pieter

-- 
Pieter Lexis
PowerDNS.COM BV -- https://www.powerdns.com