So, when a zone gets updated, PowerDNS checks if it has to notify slaves and send NOTIFYs to the slaves. Once this is done, it updates the notified_serial in the domains table. This UPDATE, for whatever reason, causes a deadlock.
Maybe you can work around that problem. You mentioned that you use DB replication. So, if all the slaves get their zones via DB replication and NOT via NOTIFY/AXFR, then you can disable the NOTIFYs and "UPDATE notified_serial...." query by setting master=no in pdns.conf. So, if Powerdns need not send NOTIFYs, then set master=no. regards Klaus Von: Pdns-users <pdns-users-boun...@mailman.powerdns.com> Im Auftrag von Klaus Darilion via Pdns-users Gesendet: Freitag, 18. Februar 2022 09:58 An: Riccardo Brunetti <riccardo.brune...@host.it>; Pdns-users@mailman.powerdns.com Betreff: Re: [Pdns-users] powerdns frequently crashes and restarts Look in the mysql-server log. Maybe it mentions which other transaction causes the deadlock. That my help to find the cause. Is this problem something new? I.e. started it after a PDNs or Mysql Upgrade? regards Klaus Von: Pdns-users <pdns-users-boun...@mailman.powerdns.com<mailto:pdns-users-boun...@mailman.powerdns.com>> Im Auftrag von Riccardo Brunetti via Pdns-users Gesendet: Freitag, 18. Februar 2022 09:55 An: pdns-users@mailman.powerdns.com<mailto:pdns-users@mailman.powerdns.com> Betreff: [Pdns-users] powerdns frequently crashes and restarts Hello. We are having quite frequent crashes of our pdns server with the following syslog messages: Feb 17 13:39:33 xxxx.xxxxx.xxx.xxxx pdns_server[26672]: Exiting because communicator thread died with error: GSQLBackend unable to refresh domain_id XXXXX: Could not execute mysql statement: update domains set notified_serial=? where id=?: Deadlock found when trying to get lock; try restarting transaction Feb 17 13:39:34 xxxx.xxxxx.xxx.xxxx systemd[1]: pdns.service: main process exited, code=exited, status=1/FAILURE Feb 17 13:39:34 xxxx.xxxxx.xxx.xxxx systemd[1]: Unit pdns.service entered failed state. Feb 17 13:39:34 xxxx.xxxxx.xxx.xxxx systemd[1]: pdns.service failed. Feb 17 13:39:35 xxxx.xxxxx.xxx.xxxx systemd[1]: pdns.service holdoff time over, scheduling restart. Feb 17 13:39:35 xxxx.xxxxx.xxx.xxxx systemd[1]: Stopped PowerDNS Authoritative Server. Feb 17 13:39:35 xxxx.xxxxx.xxx.xxxx systemd[1]: Starting PowerDNS Authoritative Server... Feb 17 13:39:35 xxxx.xxxxx.xxx.xxxx pdns_server[25899]: Loading '/usr/lib64/pdns/libgmysqlbackend.so' Feb 17 13:39:35 xxxx.xxxxx.xxx.xxxx pdns_server[25899]: This is a standalone pdns I don't know if this can be related, but just before those crashes we see messages like: Feb 17 13:39:32 xxxx.xxxxx.xxx.xxxx pdns_server[26672]: Request to queue notification for domain yyyyy.yy was processed, but no valid nameserver or ALSO-NOTIFYs found. Not notifying! Just to mention briefly our setup: we are running: PowerDNS Authoritative Server 4.5.2 (C) in a percona mysql cluster with three nodes, each node having it's own mysql and pdns server (in order to have redundancy of database and APIs). This cluster acts as a "backend master" for the real authoritative slaves, placed elsewhere, replicating the db. Can you help us to understand why we have those crashes? Thanks Riccardo
_______________________________________________ Pdns-users mailing list Pdns-users@mailman.powerdns.com https://mailman.powerdns.com/mailman/listinfo/pdns-users
