Bugzilla – Bug 616
NSD gives up on xfer
Last modified: 2014-10-10 15:47:38 CEST
When starting a secondary slave NSD server and master server is unavailable for some time after NSD starts, NSD will attempt XFR, but never again.
If the database is available, NSD will use that information if XFR fails, but only if the information is already available in the database or a zonefile exists.
Using nsd-control to reload or reconfig does not "kick" NSD to attempt the XFR again if the master is available. Only a restart of NSD will get NSD to do an initial xfer from the master server.
Enhancement suggestion: Implement an initial xfer retry timer?
NSD does retry this, but there is an exponential backoff for that timer. Wait longer, or use nsd-control transfer or force_transfer to kick it and make it do another transfer attempt.
Does that solve your problem, or is there a bug (it should really retry) ?
Ok, that is good to know. It's possible we did not wait long enough. Could some more information be documented on the exponential timer or can a knob be implemented to have more control of this behavior?
If nsd-control transfer reset this backoff to the initial timer (about 10 seconds I think), would that help? I could then document that with that command.
Thanks for the report, the fix is in the code repository.