Typical Bind slave failure scenario - What happens and when?

[Mark Andrews]

In message <503BF723.6030803 at jonesmail.me>, Russell Jones writes:
> Bind 9.7 on CentOS 6.2
> 
> 
> Hi all,
> 
> I am attempting to recreate a failure scenario of a slave server that 
> cannot contact its master for a zone. I am having a difficult time 
> following the log entries of what is occurring due to Bind not seemingly 
> following the SOA definitions for the zone. I have the following set on 
> this test zone:
> 
> example.com             IN SOA  dns1.example.com. root.dns1.example.com. (
>                                  2012071914 ; serial
>                                  60         ; refresh (1 minute)
>                                  30         ; retry (30 seconds)
>                                  120        ; expire (2 minutes)
>                                  86400      ; minimum (1 day)
>                                  )
> 
> 
> 
> I am running into the following issues:
> 
>   * The refresh time seems to be ignored. Even though it is set at 60
>     seconds, it seemed to poll the master server somewhere between 3 and
>     5 minutes. I had to define max and min refresh time in
>     /etc/named.conf to get it to respect this time. Is there some
>     internal limit to how low of a setting  Bind will allow in some of
>     the SOA definitions? I was unable to find a minimum limit for this
>     when Googling :-)

The limits are set to sane default values and are documented in
the BIND 9 Administrators Reference Manual.

		{min,max}-{retry,refresh}
 
>   * I am seeing the following messages when it fails to contact the
>     master (as expected). Only problem is after it sends this message I
>     am not seeing an attempt to contact the master every 30 seconds as I
>     am interpreting the "retry" limit to mean:
>       o zone example.com/IN: refresh: retry limit for master
>         10.32.69.1#53 exceeded (source 0.0.0.0#0)
> 
> 
> Am I misunderstanding what Bind will log and attempt to do with the 
> "retry" limit?
> 
> Is there any documentation outlining what will actually occur, and when, 
> with a slave server when it cannot contact a zone's master for updates?

It will keep retrying to refresh the zone forever.  It will stop
serving the zone when the expire timer triggers if it can't refresh
the zone.

> Thanks!
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org