bind crash with max-refresh-time 0;

[Matus UHLAR - fantomas]

>> >Does this also stop a slave from checking when it receives a
>> >notify? The documentation isn't clear on that.
>>
>> configure master not to send notifies then. Alternatively, you can
>> deny notifies from master. But the first Mark's question is still
>> important:
>> What are you trying to achieve?

On 03.02.12 11:05, Miek Gieben wrote:
>We were (are?) seeing a bug when using multiple masters. If during a zone
>transfer a notify is sent, it looks like BIND aborts the transfer and
>tries the second master. This second master is a spare standby and it
>normally turned off. When BIND hits this second master it sees it
>cannot do an axfr. BIND then (this is the bug) does not return to the
>first master to finish (or restart) the transfer. It just sits until
>the retry timer expires, which in this case is 15 minutes.
>
>We notified ISC of this, but replicating this bug was hard and we
>needed to go in production. (Sadly bind bugs aren't searchable on the
>internet).
>
>So to work around this I thought: kill the SOA timers (messing with the
>zone is not an option) and only use notifies. But then bind crashes :)

Are you sure that only xferring when NOTIFY is received will prevent 
from crashing when another NOTIFY is received during transfer triggered 
by one NOTIFY?

I doubt so. In such case, better aproach should be disabling NOTIFY and 
only transferring when timers expire. 

However, the best approach should be upgrading to 9.8 and/or trying to 
replicate the problem (using unstripped BIND with debug informations and 
inspecting core file).

-- 
Matus UHLAR - fantomas, uhlar at fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
99 percent of lawyers give the rest a bad name.