[Pdns-users] PowerDNS authoritative, galera

[Klaus Darilion]

Hi Nick!

I found this old thread. I am not familiar with galera, Percona etc. but
IMO the setup is to complex. Why do you use master-master replication
between the slaves? The slaves get synced from the master, hence every
slave can have its own DB, no clustering needed.

if both slaves receive updates at the same time, and both slaves write
to the galera-cluster at the same time, this may cause some locking.


regards
Klaus

Am 16.08.2017 um 04:17 schrieb Nick Urbanik:
> Dear Folks,
> 
> Thank you for the great software.
> 
> BACKGROUND:
> 
> We have just set up an authoritative pair of slaves at two different
> locations.  The two slaves at each location share an anycast address.
> Both slaves read/write to a galera backend through a unix socket.  We
> are using PowerDNS 4.0.4 on RHEL 7.
> 
> These slaves use ordinary DNS replication from a master with a mysql
> backend.  The slaves are using Percona-XtraDB-Cluster 5.7.17.  The
> masters are using Percona-Server 5.7.18 with corosync and pacemaker,
> DRBD.
> 
> PROBLEM:
> 
> We have PowerDNS failing quite often (about twice daily) with this message:
> Aug 16 04:11:45 dnsslave2 pdns-instance[20433]: Exiting because
> communicator thread died with error: GSQLBackend unable to refresh
> domain_id 18537: Could not
> execute mysql statement: update domains set last_check=? where id=?:
> Deadlock f
> ound when trying to get lock; try restarting transaction
> Aug 16 04:11:46 dnsslave2 pdns-instance[20433]: Got a signal 6,
> attempting to
> print trace:
> 
> [stack trace followed by restart by systemd]
> 
> I would be grateful for any suggestions on what we could do to remedy this.