[Pdns-users] pgsql db lookup failures

Charles Sprickman spork at bway.net
Sat Jul 9 21:24:11 UTC 2011


On Sat, 9 Jul 2011, Stefan Kaltenbrunner wrote:

> On 07/01/2011 02:08 AM, Charles Sprickman wrote:
>> Hello,
>>
>> I'm in the middle of testing a new powerdns/postgres setup in native
>> mode with postgres replication.  I've not run into any major issues so
>> far, but I'm seeing some odd behavior on the last server I setup.
>>
>> I started powerdns, and after that had stopped and started postgres a
>> few times to make some config changes.  It seems like powerdns is not
>> dealing with the disappearing postgres process, almost as if the backend
>> connection has not attempted a reconnect:
>>
>> Jun 30 19:46:07 ns3 pdns[1569]: Database module reported condition which
>> prevented lookup (PostgreSQL failed to execute command: FATAL:
>> terminating connection due to administrator command server closed the
>> connection unexpectedly     This probably means the server terminated
>> abnormally     before or while processing the request. ) sending out
>> servfail
>
> this indicates that somebody sent the PostgreSQL backend handing the
> connection a -TERM signal  to drop the connection, this has been caused
> in the past by broken linux kernel OOM killers or weird admin scripts
> sending random signals to random processes based on memory or cpu usage.
> It does not at all seem to relate powerdns directly at all...

Oh, it was certainly me.  I had to restart Postgres to pick up some config 
changes.  The problem I'm having is that until I restart PowerDNS, it 
doesn't reconnect when Postgres is again available.

Thanks,

Charles

>
> Stefan
>



More information about the Pdns-users mailing list