What does "deleted from unreachable cache" mean?

[Peter Olsson]

On Fri, Aug 03, 2012 at 09:13:50AM +0100, Cathy Almond wrote:
> On 02/08/12 19:00, Michael Hoskins (michoski) wrote:
> > -----Original Message-----
> > 
> > From: Peter Olsson <pol at leissner.se>
> > Date: Thursday, August 2, 2012 10:25 AM
> > To: Cathy Almond <cathya at isc.org>
> > Cc: "bind-users at lists.isc.org" <bind-users at lists.isc.org>
> > Subject: Re: What does "deleted from unreachable cache" mean?
> > 
> >> Excellent information, thanks!
> > 
> > Agreed.  I really appreciate the effort ISC has put into the KB.
> > 
> >> However, it is worrying that the master sometimes is unreachable.
> >> Is there some way I can make the slave server log, with timestamp,
> >> what zone it was trying to refresh when it failed?
> > 
> > Not sure if you've already tried, but do you have xfer logging enabled?
> > 
> > logging {
> > 
> > <snip>
> > 
> > 	channel audit_log {
> >     
> > 		file "/var/named/bind/named.log";
> > 		severity debug;
> > 		print-time yes;
> >     
> > 	};
> > 
> > 
> > <snip>
> > 
> > 	category xfer-in { audit_log; };
> > 	category xfer-out { audit_log; };
> > 	category notify { audit_log; };
> > 	category network { audit_log; };
> > 	category update { audit_log; };
> > 	// might want this to debug...
> > 	//category queries { audit_log; };
> > 
> > };
> 
> The point at which the 'unreachable' entry is cached, is logged under
> category 'xfer-in' - although it doesn't actually tell you that it's
> caching it.  Look for messages containing text "failed to connect" or
> "could not refresh".
> 
> Once the master is already in the unreachable cache, if the refresh code
> checks and finds it there, then there are several messages (different
> circumstances) that explain why a transfer isn't going to happen right
> then - and these ones all incorporate the text "unreachable (cached)".
> 
> But yesterday, I dug further into the code that's reporting "deleted
> from unreachable cache" and I'm sorry that I have to report that there
> is a bug there - the code is matching the source of the notify
> correctly, but may also mistakenly include and report on older cache
> entries that are already "deleted".
> 
> We'll fix this.  It's being tracked as bug ticket #30501.
> 
> But if you have no evidence of ongoing problems (looking at what's
> logged in category xfer-in - per my suggestions above) then you can
> safely ignore these messages. There will have been an issue at some
> point in the past, but which is now cleared.
> 
> Apologies.

I will try logging, but it's good to know that it might not
be a big problem.

Thanks!

-- 
Peter Olsson                    pol at leissner.se